diff --git a/multicontrol_cookbook_grid/_edges/rainy_edge.mp4 b/multicontrol_cookbook_grid/_edges/rainy_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..2276265049f6eea3054b7fd4f3158c9fe3281d64 --- /dev/null +++ b/multicontrol_cookbook_grid/_edges/rainy_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:205a8559f8b0acfcef45c4e8463ed108c1e501b99ebd066a399a2ca1e77ec51f +size 2078369 diff --git a/multicontrol_cookbook_grid/_edges/sunny_edge.mp4 b/multicontrol_cookbook_grid/_edges/sunny_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..c7305f097c4323213052295c20404d7d4b447801 --- /dev/null +++ b/multicontrol_cookbook_grid/_edges/sunny_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:07b80cbf9b4b177f1454ac4f07023d715c4aff13846d92db8197810a21e30fef +size 1223360 diff --git a/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4 b/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..5babd5754049392c9a6352d6ed2986c240d6941e --- /dev/null +++ b/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5da274b195348bb2aba7a21a151725caaab7e82af98890e0ea321292509e1fc1 +size 909940 diff --git a/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4 b/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..b2b5035f6d2d61e994e7c6b42a05eaca67fc7384 --- /dev/null +++ b/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9a960f90d662fd6f977275d796d8a1618338c1076b393d0c371558614b5f133 +size 379094 diff --git a/multicontrol_cookbook_grid/rainy/Building/color/config.yaml b/multicontrol_cookbook_grid/rainy/Building/color/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..274ca08105b9a474ce61f16980fa1db045f2b906 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/Building/color/config.yaml @@ -0,0 +1,882 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + keys_to_skip_loading: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'False' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataloaders: + image_data: + ratio: '0' + video_data: + dataloader: + batch_size: '1' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '32' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control_and_image_context + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' + use_cache: 'False' + ratio: '1' + video_data_1: + dataloader: + batch_size: '6' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '32' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control_and_image_context + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + num_video_frames: '1' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' + use_cache: 'False' + ratio: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_control_conditioner +- ema: power +- tokenizer: wan2pt1_tokenizer +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: vid2vid_2B_control + name: vid2vid_2B_control_720p_t24_control_layer4_cr1pt1_embedding_rectified_flow_with_image_context_with_image_data + project: cosmos_transfer2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + base_load_from: null + conditional_frame_timestep: -1.0 + conditional_frames_probs: + 0: 0.4 + 1: 0.4 + 2: 0.2 + conditioner: + _target_: + control_input_depth: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth + output_key: control_input_depth + control_input_depth_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth_mask + output_key: control_input_depth_mask + control_input_edge: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge + output_key: control_input_edge + control_input_edge_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge_mask + output_key: control_input_edge_mask + control_input_inpaint: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint + output_key: control_input_inpaint + control_input_inpaint_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint_mask + output_key: control_input_inpaint_mask + control_input_seg: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg + output_key: control_input_seg + control_input_seg_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg_mask + output_key: control_input_seg_mask + control_input_vis: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis + output_key: control_input_vis + control_input_vis_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis_mask + output_key: control_input_vis_mask + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + reference_image_context: + _target_: + dropout_rate: '0.5' + input_key: + - images + - video + - image_context + num_token: '256' + output_key: null + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + copy_weight_strategy: first_n + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + hint_keys: edge + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + condition_strategy: spaced + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_image_context_dim: '1152' + extra_per_block_abs_pos_emb: 'False' + img_context_deep_proj: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + mlp_ratio: '4.0' + model_channels: '2048' + num_blocks: '28' + num_control_branches: '1' + num_heads: '16' + num_max_modalities: '8' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + separate_embedders: 'False' + share_q_in_i2v_cross_attn: 'False' + spatial_compression_factor: '8' + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_after_proj_for_multi_branch: 'True' + use_crossattn_projection: 'True' + use_cuda_graphs: 'False' + use_input_hint_block: 'False' + use_wan_fp32_strategy: 'True' + vace_block_every_n: '7' + vace_has_mask: 'False' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: uniform + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: false + use_lora: false + use_reference_image: true + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 8 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: '8.63e-05' + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '100000' + f_max: + - '0.5' + f_min: + - '0.2' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'True' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'False' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + frame_loss_log: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '100' + hit_thres: '300' + save_s3: 'True' + save_s3_every_log_n: '10' + load_base_model: + _target_: + config: null + trainer: null + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + compile_config: + recompile_limit: '8' + use_duck_shape: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '100000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/multicontrol_cookbook_grid/rainy/Building/color/console.log b/multicontrol_cookbook_grid/rainy/Building/color/console.log new file mode 100644 index 0000000000000000000000000000000000000000..6623ba063b6a78037c83452319cc6436612ff459 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/Building/color/console.log @@ -0,0 +1,10 @@ +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:146:__init__] Found 1 hint keys across all samples +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color/config.yaml +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['rainy_Building_color'] +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample rainy_Building_color +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.json +[06-26 08:42:32|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None} +[06-26 08:53:04|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color_control_edge.mp4 +[06-26 08:53:04|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:53:05|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.mp4 diff --git a/multicontrol_cookbook_grid/rainy/Building/color/debug.log b/multicontrol_cookbook_grid/rainy/Building/color/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..3864619ebef00505b31fb3bae226226f4f6e76ea --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/Building/color/debug.log @@ -0,0 +1,969 @@ +[06-26 08:42:09|DEBUG|packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, training=True, smoke=False, verbose=False, experimental_checkpoints=False) +[06-26 08:42:10|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:42:10|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:42:21|CRITICAL|cosmos_transfer2/_src/interactive/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:42:21|CRITICAL|cosmos_transfer2/_src/predict2/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:42:22|DEBUG|cosmos_transfer2/inference.py:46:__init__] SetupArguments(output_dir=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color') model='edge' checkpoint_path='s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000' experiment='edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5' config_file='cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/config.py' context_parallel_size=1 disable_guardrails=True offload_guardrail_models=False keep_going=True profile=False benchmark=False compile_tokenizer= enable_parallel_tokenizer=False parallel_tokenizer_grid=(-1, -1))(['edge']) +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1... +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2... +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor... +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1_with_control... +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control... +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor_with_control... +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control_and_image_context... +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_with_control_input... +[06-26 08:42:22|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering hdmap_augmentor_for_local_datasets... +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package ('flash_attn_3_nv'_) was not found. +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version 0.21.0, expected at least 0.21.5.dev9. +[06-26 08:42:22|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_large_scale +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_small_overfit +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.experiment_list +[06-26 08:42:22|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.exp_large_scale +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.experiment_list +[06-26 08:42:22|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2.experiments +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot.cosmos_agibot_posttrain_example +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview.cosmos_multiview_example +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview +[06-26 08:42:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview.cosmos_singleview_example +[06-26 08:42:23|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:42:23|INFO|cosmos_transfer2/_src/imaginaire/utils/misc.py:151:set_random_seed] Using random seed 0. +[06-26 08:42:23|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:42:23|WARNING|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[06-26 08:42:23|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[06-26 08:42:23|INFO|cosmos_transfer2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/f176dc95b4a70f53ce01c4b302851595e7322b00/tokenizer.pth +[06-26 08:42:23|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.53 s +[06-26 08:42:27|INFO|cosmos_transfer2/_src/transfer2/networks/siglip2.py:65:get_siglip2_model_processor] +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #4-control_input_edge: + ReMapkey + input key: control_input_edge + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #5-control_input_vis: + ReMapkey + input key: control_input_vis + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #6-control_input_depth: + ReMapkey + input key: control_input_depth + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #7-control_input_seg: + ReMapkey + input key: control_input_seg + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #8-control_input_inpaint: + ReMapkey + input key: control_input_inpaint + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #9-control_input_edge_mask: + ReMapkey + input key: control_input_edge_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge_mask + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #10-control_input_vis_mask: + ReMapkey + input key: control_input_vis_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis_mask + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #11-control_input_depth_mask: + ReMapkey + input key: control_input_depth_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth_mask + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #12-control_input_seg_mask: + ReMapkey + input key: control_input_seg_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg_mask + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #13-control_input_inpaint_mask: + ReMapkey + input key: control_input_inpaint_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint_mask + Dtype: None +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #14-reference_image_context: + SigLip2EmbImgContext + input key: ['images', 'video', 'image_context'] + Param count: 0 + Trainable: None + Dropout rate: 0.5 + Input key: ['images', 'video', 'image_context'] + Output key: ['img_context_emb'] +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:42:27|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:27|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.09 s +[06-26 08:42:28|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.91 s +[06-26 08:42:28|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 4.42 s +[06-26 08:42:28|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 4.42 s +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:569:freeze_base_model] +Freezing base model + +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:607:freeze_base_model] ✓ Unfroze img_context_proj +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 0 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 1 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 2 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 3 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 4 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 5 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 6 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 7 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 8 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 9 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 10 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 11 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 12 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 13 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 14 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 15 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 16 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 17 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 18 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 19 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 20 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 21 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 22 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 23 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 24 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 25 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 26 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 27 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 0 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 1 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 2 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 3 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:779:load_base_model] Done loading the base model checkpoint. +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 0-th block weight to control net's 0-th block +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 1-th block weight to control net's 1-th block +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 2-th block weight to control net's 2-th block +[06-26 08:42:28|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 3-th block weight to control net's 3-th block +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[06-26 08:42:28|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[06-26 08:42:28|INFO|cosmos_transfer2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[06-26 08:42:28|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[06-26 08:42:28|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[06-26 08:42:28|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[06-26 08:42:31|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[06-26 08:42:31|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[06-26 08:42:31|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:76:__init__] MinimalV4LVGControlVaceDiT( + (x_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=72, out_features=2048, bias=False) + ) + ) + (pos_embedder): VideoRopePosition3DEmb() + (t_embedder): Sequential( + (0): Timesteps() + (1): TimestepEmbedding( + (linear_1): Linear(in_features=2048, out_features=2048, bias=False) + (activation): SiLU() + (linear_2): Linear(in_features=2048, out_features=6144, bias=False) + ) + ) + (blocks): ModuleList( + (0-27): 28 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlAwareDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + ) + ) + ) + (final_layer): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): FinalLayer( + (layer_norm): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (linear): Linear(in_features=2048, out_features=64, bias=False) + (adaln_modulation): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=4096, bias=False) + ) + ) + ) + ) + ) + (t_embedding_norm): RMSNorm() + (img_context_proj): Sequential( + (0): Linear(in_features=1152, out_features=2048, bias=True) + (1): GELU(approximate='none') + ) + (crossattn_proj): Sequential( + (0): Linear(in_features=100352, out_features=1024, bias=True) + (1): GELU(approximate='none') + ) + (control_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=520, out_features=2048, bias=False) + ) + ) + (control_blocks): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) +) +[06-26 08:42:31|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 8.40 s +[06-26 08:42:31|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 08:42:31|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 08:42:32|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:42:32|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +[06-26 08:42:32|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:271:load_model_state_dict_from_checkpoint] Checkpoint weights loaded from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt: None +[06-26 08:42:32|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:273:load_model_state_dict_from_checkpoint] Checkpoint weights loaded successfully (strict=True). +[06-26 08:42:32|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:1015:compile_tokenizer_if_enabled] Tokenizer compilation disabled +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:146:__init__] Found 1 hint keys across all samples +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color/config.yaml +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['rainy_Building_color'] +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample rainy_Building_color +[06-26 08:42:32|DEBUG|cosmos_transfer2/inference.py:186:_generate_sample] InferenceArguments(name='rainy_Building_color' prompt_path=None prompt='The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. The streets are lined with ornate classical stone buildings featuring columns, cornices and tall arched windows, their facades darkened and glistening from the rain. A wet road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, multiple vehicles move through the junction, and the camera follows the moving ego vehicle smoothly. Photorealistic, soft diffuse overcast light, high detail.' negative_prompt='The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Modern glass towers, plain concrete blocks, melted or warped facades, floating disconnected structures.' seed=2025 guidance=3 video_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4') max_frames=None context_frame_index=None image_context_path=None num_conditional_frames=1 resolution='720' sigma_max=None num_video_frames_per_chunk=93 num_steps=35 show_control_condition=False show_input=False keep_input_resolution=True edge=EdgeConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4'), control_weight=1.0, mask_path=None, mask_prompt=None, preset_edge_threshold='medium') depth=None vis=None seg=None guided_generation_mask=None guided_generation_step_threshold=25 guided_generation_foreground_labels=None) +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.json +[06-26 08:42:32|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:42:32|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None} +[06-26 08:42:32|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:472:generate_img2world] Loading input video... +[06-26 08:42:32|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4 +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:492:generate_img2world] Computing prompt text embeddings... +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:501:generate_img2world] Computing negative prompt text embeddings... +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:513:generate_img2world] Processing image context if available... +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:330:read_and_process_image_context] No image context provided. +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:525:generate_img2world] Loading control inputs... +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4 +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 1/2 +[06-26 08:42:33|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 584845 +[06-26 08:42:34|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 0 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:42:34|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:255:get_condition_with_negative_prompt] Getting condition with image context +[06-26 08:42:34|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:42:34|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:257:get_condition_with_negative_prompt] Getting uncondition with negative prompt without image context +[06-26 08:42:34|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:42:52|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:47:57|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 2/2 +[06-26 08:47:57|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 643774 +[06-26 08:47:57|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 1 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:47:57|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:255:get_condition_with_negative_prompt] Getting condition with image context +[06-26 08:47:57|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:47:57|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:257:get_condition_with_negative_prompt] Getting uncondition with negative prompt without image context +[06-26 08:47:57|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:48:00|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:53:03|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:736:generate_img2world] Average time per chunk: 313.2058851260008 +[06-26 08:53:03|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:53:04|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color_control_edge.mp4 +[06-26 08:53:04|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:53:04|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:53:05|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.mp4 diff --git a/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.json b/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.json new file mode 100644 index 0000000000000000000000000000000000000000..e4edb0ee4b59cc145094bef1170a4516e3f1bb9b --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.json @@ -0,0 +1 @@ +{"name":"rainy_Building_color","prompt_path":null,"prompt":"The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. The streets are lined with ornate classical stone buildings featuring columns, cornices and tall arched windows, their facades darkened and glistening from the rain. A wet road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, multiple vehicles move through the junction, and the camera follows the moving ego vehicle smoothly. Photorealistic, soft diffuse overcast light, high detail.","negative_prompt":"The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Modern glass towers, plain concrete blocks, melted or warped facades, floating disconnected structures.","seed":2025,"guidance":3,"video_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4","max_frames":null,"context_frame_index":null,"image_context_path":null,"num_conditional_frames":1,"resolution":"720","sigma_max":null,"num_video_frames_per_chunk":93,"num_steps":35,"show_control_condition":false,"show_input":false,"keep_input_resolution":true,"edge":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4","control_weight":1.0,"mask_path":null,"mask_prompt":null,"preset_edge_threshold":"medium"},"depth":null,"vis":null,"seg":null,"guided_generation_mask":null,"guided_generation_step_threshold":25,"guided_generation_foreground_labels":null} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.mp4 b/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..9041678e57e61eebb6aafc22235da853e651fff4 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:98bf90ea13b09d8ba738c1130c589c9c2e35cfc96871acb154b8d97306ca89ee +size 2182589 diff --git a/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.txt b/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.txt new file mode 100644 index 0000000000000000000000000000000000000000..213c470aee0671b5a985429e153c07e4efd43cd3 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color.txt @@ -0,0 +1 @@ +The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. The streets are lined with ornate classical stone buildings featuring columns, cornices and tall arched windows, their facades darkened and glistening from the rain. A wet road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, multiple vehicles move through the junction, and the camera follows the moving ego vehicle smoothly. Photorealistic, soft diffuse overcast light, high detail. \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color_control_edge.mp4 b/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color_control_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..5cdebe53fd6e5d64099ea77a2b322d3c74554a2d --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/Building/color/rainy_Building_color_control_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d1a2bd621384e9ffdfe7aeacfdc3c277e04360f64aa63573dd720fda2ac14af +size 2253955 diff --git a/multicontrol_cookbook_grid/rainy/Building/color/spec.json b/multicontrol_cookbook_grid/rainy/Building/color/spec.json new file mode 100644 index 0000000000000000000000000000000000000000..e377f2b54832490f021a4c1ae9d68da26344f171 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/Building/color/spec.json @@ -0,0 +1,13 @@ +{ + "name": "rainy_Building_color", + "prompt": "The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. The streets are lined with ornate classical stone buildings featuring columns, cornices and tall arched windows, their facades darkened and glistening from the rain. A wet road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, multiple vehicles move through the junction, and the camera follows the moving ego vehicle smoothly. Photorealistic, soft diffuse overcast light, high detail.", + "video_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4", + "guidance": 3, + "num_steps": 35, + "seed": 2025, + "edge": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4", + "control_weight": 1.0 + }, + "negative_prompt": "The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Modern glass towers, plain concrete blocks, melted or warped facades, floating disconnected structures." +} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/background/background/config.yaml b/multicontrol_cookbook_grid/rainy/background/background/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..4eef9b3568b97bfa08eeb49b76b3a85817a8c1b1 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/config.yaml @@ -0,0 +1,825 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + keys_to_skip_loading: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'False' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataset: + augmentor_name: video_basic_augmentor_v2_with_control + caption_type: t2w_qwen2p5_7b + control_input_type: edge_vis_depth_segcolor + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_control_conditioner +- ema: power +- tokenizer: wan2pt1_tokenizer +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: vid2vid_2B_control + name: vid2vid_2B_control_720p_t24_control_layer4_cr1pt1_embedding_rectified_flow + project: cosmos_transfer2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + base_load_from: null + conditional_frame_timestep: -1.0 + conditional_frames_probs: + 0: 0.4 + 1: 0.4 + 2: 0.2 + conditioner: + _target_: + control_input_depth: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth + output_key: control_input_depth + control_input_depth_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth_mask + output_key: control_input_depth_mask + control_input_edge: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge + output_key: control_input_edge + control_input_edge_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge_mask + output_key: control_input_edge_mask + control_input_inpaint: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint + output_key: control_input_inpaint + control_input_inpaint_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint_mask + output_key: control_input_inpaint_mask + control_input_seg: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg + output_key: control_input_seg + control_input_seg_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg_mask + output_key: control_input_seg_mask + control_input_vis: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis + output_key: control_input_vis + control_input_vis_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis_mask + output_key: control_input_vis_mask + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + copy_weight_strategy: first_n + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + hint_keys: edge_vis_depth_seg + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + condition_strategy: spaced + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_per_block_abs_pos_emb: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + mlp_ratio: '4.0' + model_channels: '2048' + num_blocks: '28' + num_control_branches: '4' + num_heads: '16' + num_max_modalities: '8' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + separate_embedders: 'False' + spatial_compression_factor: '8' + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_after_proj_for_multi_branch: 'False' + use_crossattn_projection: 'True' + use_cuda_graphs: 'False' + use_input_hint_block: 'False' + use_wan_fp32_strategy: 'True' + vace_block_every_n: '7' + vace_has_mask: 'False' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: reweighting + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: false + use_lora: false + use_reference_image: false + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 8 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: '8.63e-05' + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '100000' + f_max: + - '0.5' + f_min: + - '0.2' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'True' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'False' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + frame_loss_log: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '100' + hit_thres: '300' + save_s3: 'True' + save_s3_every_log_n: '10' + load_base_model: + _target_: + config: null + trainer: null + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + compile_config: + recompile_limit: '8' + use_duck_shape: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '100000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/multicontrol_cookbook_grid/rainy/background/background/console.log b/multicontrol_cookbook_grid/rainy/background/background/console.log new file mode 100644 index 0000000000000000000000000000000000000000..20e7541380a4bc9e2c2cbc4cf91479231d5e6d24 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/console.log @@ -0,0 +1,13 @@ +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:146:__init__] Found 3 hint keys across all samples +[06-26 08:28:43|WARNING|cosmos_transfer2/inference.py:148:__init__] Loading the multicontrol model. Multicontrol inference is not strictly equal to single control +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/config.yaml +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['rainy_background_background'] +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample rainy_background_background +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.json +[06-26 08:28:43|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None, 'vis': None, 'vis_mask': None, 'vis_mask_prompt': None, 'seg': '/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/semseg_front.mp4', 'seg_mask': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4', 'seg_mask_prompt': None} +[06-26 08:42:02|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_edge.mp4 +[06-26 08:42:03|INFO|cosmos_transfer2/inference.py:311:_generate_sample] vis control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_vis.mp4 +[06-26 08:42:04|INFO|cosmos_transfer2/inference.py:311:_generate_sample] seg control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_seg.mp4 +[06-26 08:42:04|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:42:06|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.mp4 diff --git a/multicontrol_cookbook_grid/rainy/background/background/debug.log b/multicontrol_cookbook_grid/rainy/background/background/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..ed1b8c6428fea47e691da71e1c576a16280ab8ff --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/debug.log @@ -0,0 +1,33584 @@ +[06-26 08:28:20|DEBUG|packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, training=True, smoke=False, verbose=False, experimental_checkpoints=False) +[06-26 08:28:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:28:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:28:31|CRITICAL|cosmos_transfer2/_src/interactive/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:28:32|CRITICAL|cosmos_transfer2/_src/predict2/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:28:32|DEBUG|cosmos_transfer2/inference.py:46:__init__] SetupArguments(output_dir=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background') model='edge' checkpoint_path='s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000' experiment='edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5' config_file='cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/config.py' context_parallel_size=1 disable_guardrails=True offload_guardrail_models=False keep_going=True profile=False benchmark=False compile_tokenizer= enable_parallel_tokenizer=False parallel_tokenizer_grid=(-1, -1))(['edge', 'vis', 'seg']) +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1... +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2... +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor... +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1_with_control... +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control... +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor_with_control... +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control_and_image_context... +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_with_control_input... +[06-26 08:28:32|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering hdmap_augmentor_for_local_datasets... +[06-26 08:28:32|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package ('flash_attn_3_nv'_) was not found. +[06-26 08:28:32|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version 0.21.0, expected at least 0.21.5.dev9. +[06-26 08:28:32|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment +[06-26 08:28:32|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:28:32|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_large_scale +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_small_overfit +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.experiment_list +[06-26 08:28:33|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.exp_large_scale +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.experiment_list +[06-26 08:28:33|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2.experiments +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot.cosmos_agibot_posttrain_example +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview.cosmos_multiview_example +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview.cosmos_singleview_example +[06-26 08:28:33|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:28:33|INFO|cosmos_transfer2/_src/imaginaire/utils/misc.py:151:set_random_seed] Using random seed 0. +[06-26 08:28:33|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:28:33|WARNING|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[06-26 08:28:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/f176dc95b4a70f53ce01c4b302851595e7322b00/tokenizer.pth +[06-26 08:28:34|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.49 s +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #4-control_input_edge: + ReMapkey + input key: control_input_edge + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #5-control_input_vis: + ReMapkey + input key: control_input_vis + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #6-control_input_depth: + ReMapkey + input key: control_input_depth + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #7-control_input_seg: + ReMapkey + input key: control_input_seg + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #8-control_input_inpaint: + ReMapkey + input key: control_input_inpaint + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #9-control_input_edge_mask: + ReMapkey + input key: control_input_edge_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge_mask + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #10-control_input_vis_mask: + ReMapkey + input key: control_input_vis_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis_mask + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #11-control_input_depth_mask: + ReMapkey + input key: control_input_depth_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth_mask + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #12-control_input_seg_mask: + ReMapkey + input key: control_input_seg_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg_mask + Dtype: None +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #13-control_input_inpaint_mask: + ReMapkey + input key: control_input_inpaint_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint_mask + Dtype: None +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:28:34|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.09 s +[06-26 08:28:34|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.81 s +[06-26 08:28:34|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 0.81 s +[06-26 08:28:34|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 0.81 s +[06-26 08:28:34|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:569:freeze_base_model] +Freezing base model + +[06-26 08:28:34|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:779:load_base_model] Done loading the base model checkpoint. +[06-26 08:28:34|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 0-th block weight to control net's 0-th block +[06-26 08:28:34|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 1-th block weight to control net's 1-th block +[06-26 08:28:34|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 2-th block weight to control net's 2-th block +[06-26 08:28:34|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 3-th block weight to control net's 3-th block +[06-26 08:28:34|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[06-26 08:28:34|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[06-26 08:28:35|INFO|cosmos_transfer2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[06-26 08:28:35|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[06-26 08:28:35|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[06-26 08:28:35|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[06-26 08:28:35|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[06-26 08:28:38|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[06-26 08:28:38|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[06-26 08:28:38|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:76:__init__] MinimalV4LVGControlVaceDiT( + (x_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=72, out_features=2048, bias=False) + ) + ) + (pos_embedder): VideoRopePosition3DEmb() + (t_embedder): Sequential( + (0): Timesteps() + (1): TimestepEmbedding( + (linear_1): Linear(in_features=2048, out_features=2048, bias=False) + (activation): SiLU() + (linear_2): Linear(in_features=2048, out_features=6144, bias=False) + ) + ) + (blocks): ModuleList( + (0-27): 28 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlAwareDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + ) + ) + ) + (final_layer): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): FinalLayer( + (layer_norm): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (linear): Linear(in_features=2048, out_features=64, bias=False) + (adaln_modulation): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=4096, bias=False) + ) + ) + ) + ) + ) + ) + ) + ) + (t_embedding_norm): RMSNorm() + (crossattn_proj): Sequential( + (0): Linear(in_features=100352, out_features=1024, bias=True) + (1): GELU(approximate='none') + ) + (control_embedder): ModuleList( + (0-3): 4 x PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=520, out_features=2048, bias=False) + ) + ) + ) + (control_blocks_0): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_1): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_2): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_3): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) +) +[06-26 08:28:38|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 4.82 s +[06-26 08:28:38|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 08:28:38|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 08:28:39|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:39|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'control_blocks.0.self_attn.q_proj.weight', 'control_blocks.0.self_attn.q_norm.weight', 'control_blocks.0.self_attn.k_proj.weight', 'control_blocks.0.self_attn.k_norm.weight', 'control_blocks.0.self_attn.v_proj.weight', 'control_blocks.0.self_attn.output_proj.weight', 'control_blocks.0.cross_attn.q_proj.weight', 'control_blocks.0.cross_attn.q_norm.weight', 'control_blocks.0.cross_attn.k_proj.weight', 'control_blocks.0.cross_attn.k_norm.weight', 'control_blocks.0.cross_attn.v_proj.weight', 'control_blocks.0.cross_attn.output_proj.weight', 'control_blocks.0.cross_attn.k_img.weight', 'control_blocks.0.cross_attn.v_img.weight', 'control_blocks.0.cross_attn.q_img.weight', 'control_blocks.0.cross_attn.q_img_norm.weight', 'control_blocks.0.cross_attn.k_img_norm.weight', 'control_blocks.0.mlp.layer1.weight', 'control_blocks.0.mlp.layer2.weight', 'control_blocks.0.adaln_modulation_self_attn.1.weight', 'control_blocks.0.adaln_modulation_self_attn.2.weight', 'control_blocks.0.adaln_modulation_cross_attn.1.weight', 'control_blocks.0.adaln_modulation_cross_attn.2.weight', 'control_blocks.0.adaln_modulation_mlp.1.weight', 'control_blocks.0.adaln_modulation_mlp.2.weight', 'control_blocks.0.before_proj.weight', 'control_blocks.0.before_proj.bias', 'control_blocks.0.after_proj.weight', 'control_blocks.0.after_proj.bias', 'control_blocks.1.self_attn.q_proj.weight', 'control_blocks.1.self_attn.q_norm.weight', 'control_blocks.1.self_attn.k_proj.weight', 'control_blocks.1.self_attn.k_norm.weight', 'control_blocks.1.self_attn.v_proj.weight', 'control_blocks.1.self_attn.output_proj.weight', 'control_blocks.1.cross_attn.q_proj.weight', 'control_blocks.1.cross_attn.q_norm.weight', 'control_blocks.1.cross_attn.k_proj.weight', 'control_blocks.1.cross_attn.k_norm.weight', 'control_blocks.1.cross_attn.v_proj.weight', 'control_blocks.1.cross_attn.output_proj.weight', 'control_blocks.1.cross_attn.k_img.weight', 'control_blocks.1.cross_attn.v_img.weight', 'control_blocks.1.cross_attn.q_img.weight', 'control_blocks.1.cross_attn.q_img_norm.weight', 'control_blocks.1.cross_attn.k_img_norm.weight', 'control_blocks.1.mlp.layer1.weight', 'control_blocks.1.mlp.layer2.weight', 'control_blocks.1.adaln_modulation_self_attn.1.weight', 'control_blocks.1.adaln_modulation_self_attn.2.weight', 'control_blocks.1.adaln_modulation_cross_attn.1.weight', 'control_blocks.1.adaln_modulation_cross_attn.2.weight', 'control_blocks.1.adaln_modulation_mlp.1.weight', 'control_blocks.1.adaln_modulation_mlp.2.weight', 'control_blocks.1.after_proj.weight', 'control_blocks.1.after_proj.bias', 'control_blocks.2.self_attn.q_proj.weight', 'control_blocks.2.self_attn.q_norm.weight', 'control_blocks.2.self_attn.k_proj.weight', 'control_blocks.2.self_attn.k_norm.weight', 'control_blocks.2.self_attn.v_proj.weight', 'control_blocks.2.self_attn.output_proj.weight', 'control_blocks.2.cross_attn.q_proj.weight', 'control_blocks.2.cross_attn.q_norm.weight', 'control_blocks.2.cross_attn.k_proj.weight', 'control_blocks.2.cross_attn.k_norm.weight', 'control_blocks.2.cross_attn.v_proj.weight', 'control_blocks.2.cross_attn.output_proj.weight', 'control_blocks.2.cross_attn.k_img.weight', 'control_blocks.2.cross_attn.v_img.weight', 'control_blocks.2.cross_attn.q_img.weight', 'control_blocks.2.cross_attn.q_img_norm.weight', 'control_blocks.2.cross_attn.k_img_norm.weight', 'control_blocks.2.mlp.layer1.weight', 'control_blocks.2.mlp.layer2.weight', 'control_blocks.2.adaln_modulation_self_attn.1.weight', 'control_blocks.2.adaln_modulation_self_attn.2.weight', 'control_blocks.2.adaln_modulation_cross_attn.1.weight', 'control_blocks.2.adaln_modulation_cross_attn.2.weight', 'control_blocks.2.adaln_modulation_mlp.1.weight', 'control_blocks.2.adaln_modulation_mlp.2.weight', 'control_blocks.2.after_proj.weight', 'control_blocks.2.after_proj.bias', 'control_blocks.3.self_attn.q_proj.weight', 'control_blocks.3.self_attn.q_norm.weight', 'control_blocks.3.self_attn.k_proj.weight', 'control_blocks.3.self_attn.k_norm.weight', 'control_blocks.3.self_attn.v_proj.weight', 'control_blocks.3.self_attn.output_proj.weight', 'control_blocks.3.cross_attn.q_proj.weight', 'control_blocks.3.cross_attn.q_norm.weight', 'control_blocks.3.cross_attn.k_proj.weight', 'control_blocks.3.cross_attn.k_norm.weight', 'control_blocks.3.cross_attn.v_proj.weight', 'control_blocks.3.cross_attn.output_proj.weight', 'control_blocks.3.cross_attn.k_img.weight', 'control_blocks.3.cross_attn.v_img.weight', 'control_blocks.3.cross_attn.q_img.weight', 'control_blocks.3.cross_attn.q_img_norm.weight', 'control_blocks.3.cross_attn.k_img_norm.weight', 'control_blocks.3.mlp.layer1.weight', 'control_blocks.3.mlp.layer2.weight', 'control_blocks.3.adaln_modulation_self_attn.1.weight', 'control_blocks.3.adaln_modulation_self_attn.2.weight', 'control_blocks.3.adaln_modulation_cross_attn.1.weight', 'control_blocks.3.adaln_modulation_cross_attn.2.weight', 'control_blocks.3.adaln_modulation_mlp.1.weight', 'control_blocks.3.adaln_modulation_mlp.2.weight', 'control_blocks.3.after_proj.weight', 'control_blocks.3.after_proj.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_embedder.proj.1.weight'], incorrect_shapes=[]) +[06-26 08:28:39|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:271:load_model_state_dict_from_checkpoint] Checkpoint weights loaded from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt: None +[06-26 08:28:39|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:273:load_model_state_dict_from_checkpoint] Checkpoint weights loaded successfully (strict=True). +[06-26 08:28:39|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 08:28:39|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 0 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 08:28:40|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0148, 0.0025, 0.0065], + [-0.0004, 0.0057, -0.0327, ..., -0.0078, 0.0021, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0136, -0.0035, 0.0283], + ..., + [-0.0028, 0.0125, 0.0248, ..., 0.0437, -0.0069, -0.0192], + [ 0.0277, 0.0146, 0.0031, ..., -0.0028, -0.0090, 0.0146], + [ 0.0437, 0.0366, 0.0082, ..., 0.0029, -0.0126, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0044, -0.0310, 0.0300, ..., -0.0064, -0.0045, -0.0339], + [-0.0103, -0.0047, 0.0114, ..., 0.0052, 0.0092, 0.0520], + [-0.0061, 0.0131, 0.0398, ..., -0.0096, -0.0179, -0.0056], + ..., + [-0.0048, 0.0175, -0.0101, ..., 0.0225, 0.0043, 0.0042], + [-0.0209, -0.0081, 0.0089, ..., -0.0308, 0.0108, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0055, 0.0354, -0.0082, ..., 0.0120, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0013, ..., 0.0376, 0.0532, 0.0036], + [ 0.0175, -0.0347, 0.0013, ..., -0.0066, 0.0053, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0154, 0.0262, -0.0112], + [-0.0006, 0.0244, 0.0356, ..., -0.0036, 0.0170, 0.0118], + [ 0.0486, 0.0175, 0.0042, ..., 0.0036, -0.0026, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0114, ..., -0.0070, -0.0337, 0.0077], + [ 0.0491, 0.0413, 0.0251, ..., -0.0030, 0.0393, 0.0042], + [-0.0162, -0.0019, 0.0140, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0038, 0.0261, 0.0039, ..., -0.0040, 0.0347, 0.0009], + [-0.0254, 0.0063, 0.0019, ..., 0.0093, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0121, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0035, -0.0270, -0.0166, ..., 0.0115, -0.0298, 0.0532], + [ 0.0017, 0.0040, -0.0131, ..., -0.0161, 0.0146, 0.0308], + [ 0.0146, 0.0087, 0.0143, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0064, -0.0630, -0.0048, ..., 0.0074, -0.0183, 0.0217], + [ 0.0023, -0.0035, 0.0258, ..., -0.0486, -0.0349, -0.0059], + [ 0.0182, -0.0076, 0.0197, ..., 0.0142, -0.0226, -0.0212]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0051, -0.0012, -0.0011, ..., -0.0005, -0.0018, -0.0120], + [-0.0153, 0.0255, 0.0432, ..., -0.0045, 0.0049, -0.0069], + [-0.0154, -0.0206, -0.0245, ..., -0.0119, -0.0325, -0.0013], + ..., + [-0.0034, 0.0039, 0.0417, ..., 0.0280, 0.0036, -0.0047], + [-0.0054, 0.0041, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0055, -0.0024, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 1.5503e-02, -6.5613e-03, -4.8096e-02, ..., 4.4922e-02, + 2.1240e-02, 1.2756e-02], + [ 1.5381e-02, -2.2705e-02, 2.1973e-02, ..., -2.2095e-02, + 3.5858e-03, 6.9885e-03], + [-1.5259e-02, -1.0071e-02, 4.2969e-02, ..., -9.5825e-03, + 3.2715e-02, -1.1169e-02], + ..., + [ 3.1006e-02, -3.7598e-02, -4.3457e-02, ..., -2.2461e-02, + 2.5513e-02, 1.3428e-02], + [-3.8862e-05, 1.6602e-02, -1.6174e-03, ..., -4.0283e-03, + 6.7139e-03, -9.6436e-03], + [-1.0864e-02, -3.4637e-03, 1.1780e-02, ..., 4.6997e-03, + -3.7994e-03, -7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[ 0.0049, -0.0386, -0.0112, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0079, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0127, 0.0032, -0.0114, ..., 0.0214, 0.0110, -0.0045], + ..., + [-0.0155, -0.0172, -0.0136, ..., 0.0095, 0.0098, 0.0184], + [-0.0151, -0.0356, 0.0200, ..., -0.0060, -0.0112, 0.0332], + [-0.0396, -0.0120, -0.0089, ..., -0.0284, -0.0140, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 2.7710e-02, 3.1250e-02, 1.3306e-02, ..., -1.9531e-02, + 1.1414e-02, -1.4954e-02], + [ 1.9165e-02, 9.2773e-03, -7.5378e-03, ..., -1.7853e-03, + -1.2695e-02, -1.9775e-02], + [-2.5024e-02, 1.1108e-02, -1.3733e-02, ..., 1.0132e-02, + 1.7212e-02, 7.0496e-03], + ..., + [ 4.6631e-02, 6.4392e-03, -1.5869e-03, ..., -4.0283e-03, + 2.3926e-02, 2.5330e-03], + [ 5.4688e-02, -1.8677e-02, -1.1108e-02, ..., -3.7537e-03, + 3.1128e-02, -5.4626e-03], + [ 8.7738e-05, 1.0742e-02, -9.6436e-03, ..., 2.0020e-02, + 1.6968e-02, 1.4038e-02]], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[-5.6028e-05, 2.3651e-03, 2.8564e-02, ..., -1.0376e-02, + 4.5471e-03, -1.8188e-02], + [ 6.2866e-03, 1.3428e-02, -3.1494e-02, ..., -4.5654e-02, + -5.0049e-03, 1.0376e-02], + [ 4.3945e-03, 2.6123e-02, 2.3926e-02, ..., 5.0049e-03, + 6.6833e-03, 4.6143e-02], + ..., + [ 1.6968e-02, -2.5391e-02, -4.3945e-02, ..., 1.5259e-02, + 4.9133e-03, -7.4768e-03], + [ 5.7678e-03, -3.7109e-02, 3.5645e-02, ..., -8.3008e-03, + 4.1016e-02, -1.7822e-02], + [ 1.7822e-02, -1.4648e-02, 5.6152e-02, ..., -2.9053e-02, + -2.6733e-02, 2.9663e-02]], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0067, 0.0208, 0.0278, ..., 0.0133, 0.0156, 0.0074], + [-0.0063, 0.0040, 0.0014, ..., 0.0135, 0.0040, 0.0045], + [ 0.0058, 0.0041, 0.0243, ..., -0.0123, -0.0049, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0064, ..., 0.0250, -0.0107, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., 0.0048, -0.0295, 0.0400], + [-0.0116, 0.0138, 0.0305, ..., -0.0103, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0054, 0.0381, -0.0239, ..., 0.0219, 0.0483, 0.0339], + [-0.0143, -0.0033, 0.0288, ..., 0.0135, 0.0583, 0.0049], + [-0.0280, 0.0154, -0.0386, ..., -0.0444, -0.0092, -0.0026], + ..., + [-0.0038, 0.0157, 0.0217, ..., 0.0048, 0.0114, 0.0154], + [ 0.0081, 0.0195, -0.0283, ..., -0.0060, -0.0021, -0.0342], + [-0.0166, -0.0195, 0.0027, ..., -0.0066, 0.0165, 0.0130]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0138, -0.0146, 0.0459, ..., -0.0028, -0.0320, 0.0277], + [-0.0026, 0.0068, 0.0270, ..., -0.0220, -0.0183, -0.0271], + [ 0.0221, 0.0219, -0.0201, ..., 0.0060, 0.0064, -0.0031], + ..., + [-0.0258, 0.0016, 0.0215, ..., 0.0038, 0.0237, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0055], + [-0.0006, -0.0080, -0.0120, ..., -0.0334, 0.0047, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0147, -0.0454, ..., -0.0132, -0.0125, 0.0400], + [ 0.0295, -0.0187, -0.0110, ..., -0.0283, 0.0047, -0.0186], + [-0.0007, 0.0547, 0.0024, ..., 0.0116, 0.0400, -0.0214], + ..., + [ 0.0280, -0.0058, -0.0266, ..., -0.0084, -0.0287, 0.0150], + [ 0.0090, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0095, 0.0102, ..., -0.0070, 0.0703, -0.0043]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0205, 0.0087, -0.0315, ..., 0.0081, 0.0254, 0.0359], + [ 0.0097, -0.0032, 0.0038, ..., -0.0120, -0.0233, -0.0048], + [ 0.0051, 0.0040, -0.0121, ..., -0.0117, -0.0070, 0.0086], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0040], + [ 0.0339, -0.0064, 0.0223, ..., 0.0410, -0.0017, 0.0136], + [-0.0046, -0.0439, -0.0486, ..., 0.0145, -0.0130, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0046, 0.0139, -0.0030, ..., -0.0081, -0.0033, -0.0229], + [-0.0289, 0.0231, 0.0077, ..., -0.0135, -0.0024, 0.0222], + [ 0.0083, 0.0295, -0.0189, ..., -0.0168, 0.0311, -0.0261], + ..., + [ 0.0233, -0.0045, -0.0036, ..., 0.0032, 0.0278, 0.0049], + [-0.0106, -0.0038, 0.0140, ..., -0.0175, 0.0444, 0.0425], + [ 0.0069, -0.0078, 0.0102, ..., 0.0066, -0.0344, 0.0229]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[-0.0050, 0.0128, 0.0190, ..., -0.0093, -0.0198, 0.0149], + [-0.0581, 0.0294, -0.0047, ..., -0.0240, 0.0043, 0.0300], + [ 0.0085, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0035, -0.0289, 0.0537, ..., 0.0058, -0.0074, -0.0099], + [-0.0029, -0.0069, 0.0210, ..., -0.0021, 0.0091, 0.0186], + [-0.0086, -0.0349, -0.0173, ..., -0.0400, 0.0052, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0177, -0.0261, -0.0112, ..., 0.0110, -0.0130, -0.0405], + [ 0.0058, -0.0086, 0.0266, ..., -0.0400, 0.0073, 0.0391], + [-0.0013, 0.0359, 0.0284, ..., 0.0286, -0.0234, 0.0217], + ..., + [ 0.0017, 0.0123, 0.0132, ..., -0.0138, 0.0032, 0.0096], + [ 0.0299, -0.0167, -0.0022, ..., -0.0063, 0.0043, -0.0206], + [ 0.0074, -0.0076, -0.0043, ..., -0.0047, -0.0067, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, -0.0032, 0.0121, ..., -0.0075, -0.0208, -0.0058], + [-0.0092, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0229], + [ 0.0132, -0.0278, 0.0115, ..., 0.0396, -0.0082, -0.0074], + ..., + [-0.0015, 0.0018, -0.0309, ..., 0.0344, 0.0073, -0.0126], + [-0.0236, -0.0102, -0.0056, ..., 0.0332, -0.0117, -0.0171], + [ 0.0231, 0.0405, 0.0103, ..., -0.0278, -0.0139, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[-0.0046, 0.0166, 0.0078, ..., 0.0273, -0.0352, 0.0153], + [-0.0510, -0.0070, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0077, ..., -0.0388, -0.0047, 0.0398], + ..., + [-0.0160, -0.0195, -0.0051, ..., -0.0040, -0.0055, -0.0035], + [-0.0187, -0.0040, 0.0097, ..., 0.0287, -0.0122, -0.0099], + [-0.0181, -0.0013, 0.0247, ..., -0.0145, 0.0072, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0078, 0.0092, 0.0140, ..., -0.0121, 0.0045, -0.0253], + [-0.0164, 0.0330, -0.0171, ..., -0.0033, -0.0403, -0.0111], + [ 0.0085, 0.0347, 0.0275, ..., 0.0151, 0.0069, 0.0080], + ..., + [ 0.0275, -0.0168, -0.0106, ..., -0.0099, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0056, ..., -0.0023, -0.0022, -0.0243], + [-0.0060, 0.0581, 0.0061, ..., 0.0148, 0.0239, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0121, -0.0410, 0.0128, ..., -0.0254, -0.0036, -0.0173], + [ 0.0476, 0.0315, 0.0141, ..., -0.0154, -0.0073, 0.0048], + [-0.0092, -0.0270, 0.0092, ..., 0.0332, 0.0150, -0.0073], + ..., + [-0.0076, 0.0039, 0.0162, ..., -0.0177, 0.0171, -0.0133], + [ 0.0044, -0.0187, -0.0051, ..., 0.0248, -0.0145, -0.0073], + [-0.0143, 0.0034, 0.0149, ..., -0.0047, -0.0006, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0049, 0.0097, ..., 0.0255, -0.0253, -0.0136], + [-0.0038, -0.0141, -0.0160, ..., -0.0109, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0117, 0.0056, 0.0322, ..., -0.0183, -0.0041, -0.0062], + [ 0.0045, -0.0054, 0.0164, ..., 0.0036, 0.0255, 0.0024], + [ 0.0270, 0.0084, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0077, ..., -0.0532, 0.0106, 0.0036], + [-0.0240, -0.0215, -0.0104, ..., -0.0063, 0.0097, -0.0168], + [ 0.0075, 0.0064, -0.0055, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0059, 0.0172, -0.0106, ..., -0.0300, -0.0005, 0.0435], + [ 0.0053, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0061], + [-0.0167, 0.0128, 0.0101, ..., -0.0247, 0.0119, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0065, -0.0079, 0.0317, ..., 0.0200, -0.0105, -0.0376], + [ 0.0356, 0.0378, 0.0129, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., 0.0076, 0.0065, -0.0322], + ..., + [ 0.0065, 0.0107, -0.0029, ..., 0.0366, -0.0206, 0.0374], + [-0.0034, -0.0079, -0.0044, ..., -0.0095, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0048, ..., -0.0166, -0.0042, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0078, -0.0096, -0.0344, ..., -0.0003, 0.0347, -0.0181], + [ 0.0267, -0.0366, -0.0089, ..., 0.0217, 0.0349, -0.0070], + [-0.0098, -0.0286, 0.0062, ..., 0.0060, -0.0282, 0.0047], + ..., + [-0.0124, -0.0039, 0.0065, ..., -0.0084, 0.0165, 0.0074], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0134], + [-0.0140, 0.0049, -0.0170, ..., -0.0126, 0.0060, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0033, 0.0051, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0120, 0.0374, ..., -0.0153, -0.0188, -0.0137], + ..., + [ 0.0262, 0.0052, -0.0065, ..., 0.0007, 0.0038, -0.0165], + [-0.0222, -0.0155, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0125, -0.0148, 0.0038, ..., 0.0018, -0.0234, 0.0017]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0102], + [ 0.0249, 0.0205, 0.0054, ..., -0.0046, -0.0093, -0.0106], + [-0.0220, -0.0004, -0.0037, ..., 0.0306, 0.0381, -0.0044], + ..., + [ 0.0200, -0.0050, 0.0170, ..., 0.0116, 0.0320, -0.0157], + [ 0.0349, 0.0119, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0049, 0.0256, 0.0045, ..., 0.0205, 0.0002, -0.0105]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0227, 0.0065, 0.0199, ..., -0.0057, 0.0276, -0.0312], + [ 0.0178, -0.0059, 0.0103, ..., 0.0112, -0.0398, 0.0293], + [ 0.0430, -0.0130, 0.0072, ..., 0.0040, -0.0069, -0.0029], + ..., + [ 0.0192, -0.0112, -0.0112, ..., 0.0027, 0.0330, 0.0137], + [-0.0125, -0.0542, -0.0205, ..., 0.0044, -0.0159, 0.0320], + [ 0.0303, 0.0047, -0.0425, ..., 0.0032, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0060, -0.0315, ..., 0.0525, 0.0041, -0.0349], + [-0.0036, -0.0026, -0.0156, ..., 0.0048, -0.0240, 0.0305], + [-0.0256, 0.0064, 0.0065, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0065, 0.0231, 0.0209, ..., 0.0076, -0.0041, 0.0461], + [-0.0063, -0.0079, -0.0019, ..., -0.0217, 0.0347, -0.0040], + [ 0.0050, 0.0234, 0.0031, ..., 0.0261, -0.0052, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0052, 0.0046], + [-0.0183, -0.0211, -0.0052, ..., -0.0149, 0.0139, -0.0162], + [-0.0078, -0.0160, 0.0079, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0096, 0.0073, ..., 0.0171, 0.0036, -0.0033], + [-0.0144, 0.0260, -0.0087, ..., 0.0041, 0.0106, -0.0166], + [ 0.0098, -0.0140, -0.0014, ..., 0.0057, 0.0054, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0078, 0.0134, -0.0064, ..., -0.0278, 0.0240, -0.0123], + [-0.0069, 0.0188, -0.0160, ..., -0.0317, -0.0064, 0.0181], + [-0.0042, 0.0310, -0.0085, ..., -0.0226, 0.0017, 0.0334], + ..., + [-0.0081, 0.0186, -0.0073, ..., 0.0228, 0.0047, 0.0064], + [ 0.0271, 0.0118, 0.0354, ..., -0.0075, 0.0198, 0.0150], + [ 0.0282, 0.0171, 0.0317, ..., -0.0109, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-1.1475e-02, -2.6489e-02, -1.0986e-02, ..., 4.8523e-03, + 3.7231e-03, 1.3428e-02], + [-3.3203e-02, -4.8523e-03, 3.0640e-02, ..., 3.4668e-02, + -3.6133e-02, 8.4839e-03], + [-5.8651e-05, -1.9043e-02, -3.6774e-03, ..., 1.6602e-02, + -1.3550e-02, 1.3733e-02], + ..., + [-2.0020e-02, -2.9144e-03, 2.4414e-04, ..., -3.3936e-02, + 3.6133e-02, 3.0518e-03], + [ 6.6528e-03, -2.5024e-02, 1.5564e-02, ..., 2.2430e-03, + -1.6235e-02, 2.1973e-02], + [-7.4463e-03, 3.8452e-03, 2.0508e-02, ..., -6.5002e-03, + -8.3008e-03, 4.3457e-02]], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0047, 0.0240, -0.0217, ..., 0.0172, 0.0046, 0.0114], + [-0.0386, 0.0106, 0.0078, ..., -0.0060, 0.0322, 0.0086], + [-0.0427, 0.0011, -0.0045, ..., 0.0076, 0.0112, -0.0137], + ..., + [-0.0087, -0.0037, -0.0264, ..., -0.0586, -0.0630, -0.0131], + [-0.0006, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0107, -0.0610, 0.0162, ..., -0.0291, -0.0166, 0.0029]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0117, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0112], + [ 0.0111, -0.0098, -0.0327, ..., 0.0447, 0.0076, 0.0059], + [ 0.0042, -0.0175, 0.0076, ..., -0.0069, -0.0037, -0.0060], + ..., + [ 0.0295, 0.0042, -0.0102, ..., 0.0099, -0.0160, -0.0286], + [ 0.0082, 0.0253, -0.0177, ..., 0.0095, 0.0064, 0.0078], + [-0.0132, -0.0048, 0.0139, ..., -0.0125, 0.0277, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0129, 0.0040, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0050, 0.0234, 0.0069, ..., 0.0464, 0.0056, 0.0309], + ..., + [-0.0118, -0.0175, 0.0327, ..., 0.0132, 0.0051, 0.0141], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, 0.0040, -0.0444, ..., 0.0060, -0.0283, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0081, -0.0100, 0.0146, ..., -0.0742, -0.0074, 0.0161], + [-0.0064, -0.0132, -0.0071, ..., 0.0053, 0.0042, 0.0238], + [ 0.0189, 0.0221, -0.0087, ..., 0.0033, 0.0408, 0.0064], + ..., + [ 0.0047, 0.0247, 0.0396, ..., 0.0032, -0.0304, -0.0183], + [ 0.0299, -0.0113, -0.0417, ..., -0.0052, -0.0376, 0.0032], + [-0.0210, 0.0168, -0.0039, ..., 0.0052, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0056, 0.0242, -0.0079, ..., -0.0245, 0.0050, -0.0057], + [ 0.0190, -0.0123, 0.0332, ..., -0.0143, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0053, ..., 0.0056, -0.0091, 0.0182], + ..., + [-0.0430, -0.0403, -0.0086, ..., 0.0134, -0.0140, 0.0060], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0089, 0.0378], + [ 0.0364, 0.0486, -0.0120, ..., -0.0060, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0094, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0078, ..., 0.0101, -0.0223, -0.0052], + ..., + [ 0.0192, 0.0292, -0.0022, ..., 0.0176, -0.0197, 0.0076], + [ 0.0265, 0.0064, 0.0057, ..., 0.0286, 0.0046, -0.0267], + [-0.0184, -0.0088, -0.0045, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0083, -0.0159, 0.0082, ..., 0.0198, 0.0044, 0.0277], + [ 0.0172, 0.0037, -0.0053, ..., 0.0054, -0.0044, 0.0074], + [-0.0019, 0.0184, 0.0113, ..., 0.0099, -0.0054, -0.0077], + ..., + [ 0.0061, -0.0111, -0.0028, ..., -0.0095, -0.0038, 0.0123], + [-0.0427, -0.0391, -0.0193, ..., 0.0096, -0.0238, -0.0078], + [ 0.0040, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0217, -0.0226, -0.0349, ..., -0.0079, -0.0161, 0.0127], + [-0.0049, 0.0008, 0.0312, ..., 0.0055, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0076, 0.0004, 0.0278], + ..., + [ 0.0170, -0.0082, 0.0471, ..., 0.0018, 0.0083, -0.0049], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0079, -0.0203], + [ 0.0033, 0.0081, -0.0077, ..., -0.0098, -0.0076, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[-0.0044, -0.0151, 0.0284, ..., -0.0109, 0.0140, -0.0002], + [ 0.0156, 0.0332, 0.0337, ..., 0.0223, -0.0165, -0.0015], + [-0.0071, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0068, 0.0062, 0.0280, ..., -0.0066, -0.0072, 0.0137], + [-0.0073, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0204], + [-0.0206, 0.0159, -0.0091, ..., 0.0045, 0.0167, -0.0086]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0055, ..., -0.0238, -0.0102, 0.0119], + [-0.0250, -0.0162, 0.0019, ..., -0.0084, 0.0035, -0.0369], + [ 0.0004, 0.0141, 0.0061, ..., -0.0452, -0.0056, 0.0121], + ..., + [ 0.0147, -0.0140, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0099, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0141, -0.0048, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[ 0.0013, -0.0327, -0.0190, ..., 0.0053, -0.0018, -0.0209], + [ 0.0713, 0.0150, -0.0026, ..., 0.0073, 0.0186, -0.0029], + [-0.0227, -0.0090, 0.0188, ..., -0.0125, 0.0193, -0.0049], + ..., + [ 0.0117, 0.0464, 0.0079, ..., 0.0248, 0.0168, -0.0038], + [ 0.0381, 0.0045, 0.0067, ..., 0.0048, -0.0110, 0.0035], + [ 0.0292, 0.0055, 0.0209, ..., 0.0044, 0.0234, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0063, 0.0212, ..., 0.0305, 0.0137, 0.0104], + [-0.0072, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0050], + ..., + [ 0.0082, -0.0275, 0.0225, ..., 0.0051, -0.0189, -0.0104], + [-0.0125, -0.0051, -0.0041, ..., 0.0046, -0.0048, 0.0211], + [ 0.0046, -0.0064, -0.0060, ..., -0.0109, 0.0264, 0.0047]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[-0.0002, -0.0076, -0.0109, ..., 0.0186, -0.0216, 0.0062], + [ 0.0045, 0.0123, -0.0231, ..., 0.0125, -0.0114, -0.0183], + [ 0.0107, 0.0103, -0.0038, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0121, 0.0146, ..., 0.0540, 0.0415, 0.0096], + [ 0.0061, 0.0084, -0.0143, ..., 0.0060, -0.0189, -0.0237], + [ 0.0048, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0067]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0019, 0.0359, ..., -0.0083, -0.0123, -0.0050], + [ 0.0190, -0.0044, 0.0298, ..., 0.0127, 0.0042, -0.0200], + [ 0.0114, -0.0112, 0.0177, ..., -0.0248, 0.0229, 0.0025], + ..., + [-0.0082, -0.0049, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0042, -0.0291, ..., 0.0206, 0.0304, -0.0089], + [-0.0081, 0.0245, -0.0066, ..., -0.0413, 0.0216, -0.0117]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0078, -0.0157, 0.0077, ..., -0.0049, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0064], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, 0.0038], + ..., + [-0.0036, 0.0003, 0.0258, ..., -0.0310, 0.0091, -0.0361], + [-0.0078, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [ 0.0067, 0.0157, -0.0070, ..., 0.0040, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0077, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0106, ..., 0.0103, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0041, ..., 0.0129, -0.0046, 0.0085], + ..., + [-0.0225, 0.0337, -0.0085, ..., -0.0177, 0.0542, -0.0249], + [ 0.0013, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0128, 0.0206, ..., 0.0071, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0095, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0052, 0.0063, ..., 0.0039, 0.0113, 0.0598], + [-0.0068, 0.0267, 0.0054, ..., 0.0284, -0.0060, -0.0364], + ..., + [ 0.0044, 0.0040, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0046, -0.0089, ..., 0.0211, 0.0135, -0.0075], + [ 0.0101, 0.0236, -0.0206, ..., -0.0135, -0.0048, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0067, 0.0294, ..., 0.0050, 0.0270, 0.0063], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0065, -0.0168], + [ 0.0201, -0.0120, 0.0126, ..., -0.0381, -0.0247, -0.0076], + ..., + [ 0.0275, -0.0182, -0.0183, ..., 0.0464, 0.0091, -0.0077], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0067, -0.0466], + [ 0.0327, 0.0284, -0.0030, ..., 0.0143, 0.0142, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, -0.0027, -0.0167, ..., -0.0039, 0.0170, 0.0107], + [-0.0002, -0.0002, 0.0295, ..., -0.0172, 0.0048, 0.0135], + [ 0.0193, 0.0236, -0.0123, ..., -0.0007, -0.0369, 0.0239], + ..., + [ 0.0150, -0.0413, -0.0126, ..., -0.0354, 0.0052, -0.0320], + [-0.0265, -0.0078, -0.0327, ..., -0.0070, 0.0077, 0.0074], + [ 0.0299, 0.0356, -0.0192, ..., 0.0059, -0.0052, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0013, 0.0041, 0.0317, ..., 0.0042, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0076, -0.0093], + [-0.0166, 0.0072, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0019, 0.0260, ..., 0.0159, 0.0048, -0.0474], + [-0.0056, 0.0080, 0.0077, ..., 0.0146, 0.0557, 0.0048], + [-0.0203, 0.0273, 0.0047, ..., -0.0330, 0.0065, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0058, ..., -0.0014, -0.0093, -0.0277], + [ 0.0060, -0.0078, -0.0062, ..., 0.0036, 0.0140, -0.0161], + [-0.0052, 0.0168, -0.0059, ..., -0.0092, 0.0124, 0.0043], + ..., + [ 0.0398, 0.0452, 0.0251, ..., 0.0039, -0.0330, 0.0051], + [-0.0620, -0.0039, 0.0371, ..., 0.0410, 0.0061, -0.0173], + [ 0.0127, 0.0074, -0.0270, ..., 0.0081, 0.0388, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0040, 0.0245], + [-0.0449, 0.0025, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0064, ..., 0.0046, 0.0288, 0.0125], + [ 0.0330, 0.0038, 0.0057, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0069, 0.0045, ..., -0.0077, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0070, 0.0038, 0.0527], + [-0.0051, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0095, -0.0096, 0.0327], + [ 0.0146, 0.0116, -0.0030, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0129, 0.0281, ..., -0.0242, -0.0126, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0155, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0041], + [-0.0048, -0.0045, -0.0286, ..., 0.0050, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0057, ..., -0.0019, 0.0082, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0096, 0.0150], + [-0.0071, 0.0029, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0063, ..., 0.0092, -0.0114, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0070, -0.0033, ..., -0.0220, -0.0145, -0.0046], + [-0.0166, 0.0189, -0.0058, ..., 0.0096, 0.0033, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0067, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0066, ..., 0.0024, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0069, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0121, 0.0289, 0.0126, ..., 0.0134, 0.0066, -0.0103], + [ 0.0188, -0.0197, -0.0022, ..., 0.0284, -0.0342, -0.0058], + [-0.0525, -0.0214, 0.0154, ..., -0.0270, 0.0081, 0.0254], + ..., + [-0.0052, 0.0087, 0.0281, ..., 0.0129, -0.0042, 0.0049], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0037, -0.0054], + [-0.0273, -0.0747, 0.0208, ..., -0.0042, -0.0052, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 9.1171e-04, 1.4954e-02, 3.3447e-02, ..., 2.0508e-02, + -2.7344e-02, -3.7537e-03], + [-1.6479e-02, 1.1108e-02, -1.3489e-02, ..., 1.0620e-02, + -7.2327e-03, 2.1851e-02], + [ 5.0659e-03, 2.2705e-02, -5.6458e-03, ..., 1.2817e-02, + -7.1106e-03, 1.4954e-02], + ..., + [-3.4668e-02, -1.4709e-02, 6.8359e-03, ..., 5.5847e-03, + -5.0659e-03, -9.2773e-03], + [-6.5613e-03, 3.9795e-02, -3.4424e-02, ..., 2.4536e-02, + 6.3477e-02, -1.1780e-02], + [-5.3024e-04, -5.0049e-02, -2.3193e-02, ..., 3.1982e-02, + -7.5340e-05, -5.5847e-03]], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0154, ..., -0.0344, 0.0067, -0.0364], + [ 0.0220, -0.0030, -0.0131, ..., 0.0078, 0.0101, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, 0.0072, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0073, 0.0192, -0.0393], + [-0.0105, -0.0244, 0.0275, ..., -0.0052, 0.0042, 0.0081], + [-0.0170, -0.0427, 0.0142, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0046, -0.0378, -0.0359, ..., 0.0552, -0.0078, -0.0160], + [ 0.0045, 0.0140, 0.0133, ..., 0.0050, -0.0153, 0.0354], + [ 0.0095, -0.0042, 0.0045, ..., -0.0047, -0.0249, -0.0136], + ..., + [ 0.0079, -0.0125, 0.0051, ..., -0.0095, -0.0084, 0.0052], + [-0.0060, 0.0439, 0.0166, ..., 0.0248, 0.0132, -0.0187], + [ 0.0175, 0.0063, 0.0044, ..., 0.0410, -0.0096, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0105, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0145, 0.0025, ..., 0.0065, -0.0244, 0.0356], + [ 0.0079, 0.0165, -0.0200, ..., -0.0080, 0.0237, -0.0081], + ..., + [-0.0197, -0.0022, 0.0386, ..., 0.0181, 0.0113, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, -0.0041, -0.0374], + [-0.0086, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0075, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0137, 0.0115, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, 0.0013, -0.0066, ..., 0.0050, -0.0045, 0.0337], + ..., + [ 0.0422, 0.0054, 0.0352, ..., -0.0304, -0.0035, -0.0204], + [ 0.0302, 0.0156, 0.0100, ..., -0.0049, -0.0254, -0.0245], + [-0.0067, -0.0064, 0.0078, ..., -0.0078, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0044, 0.0266, ..., 0.0154, 0.0187, 0.0049], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0096], + [-0.0189, 0.0205, 0.0044, ..., -0.0172, -0.0090, 0.0231], + ..., + [-0.0295, -0.0137, 0.0045, ..., 0.0039, -0.0050, 0.0099], + [-0.0052, -0.0140, 0.0317, ..., -0.0153, -0.0459, -0.0065], + [-0.0041, -0.0051, -0.0259, ..., -0.0040, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0059, -0.0140, ..., 0.0242, 0.0055, -0.0386], + [-0.0141, 0.0060, -0.0097, ..., -0.0369, -0.0155, 0.0157], + [-0.0267, -0.0041, -0.0065, ..., -0.0127, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0029, 0.0144, ..., 0.0115, 0.0040, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., 0.0045, -0.0038, -0.0137], + [-0.0219, -0.0396, -0.0010, ..., -0.0077, -0.0033, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0051, 0.0168, 0.0194, ..., 0.0349, -0.0420, -0.0048], + [ 0.0044, 0.0273, 0.0146, ..., -0.0184, 0.0011, 0.0156], + [ 0.0145, -0.0031, -0.0070, ..., 0.0245, 0.0359, 0.0061], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0037, 0.0043, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0072, -0.0312, -0.0457], + [ 0.0055, 0.0039, -0.0058, ..., -0.0038, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0047, 0.0121, 0.0038, ..., 0.0124, 0.0006, 0.0067], + [-0.0277, -0.0039, -0.0181, ..., 0.0005, 0.0020, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0038, 0.0513, -0.0113], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0113], + [-0.0282, -0.0040, 0.0121, ..., -0.0057, -0.0039, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, 0.0048, -0.0128, ..., 0.0140, 0.0308, -0.0066], + [ 0.0120, -0.0332, -0.0320, ..., -0.0045, -0.0140, -0.0039], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0088, 0.0295, 0.0505], + [-0.0159, 0.0025, -0.0311, ..., 0.0120, 0.0063, 0.0081], + [ 0.0044, 0.0198, 0.0141, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, 0.0043, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0090, -0.0040, 0.0052, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0045, ..., 0.0302, -0.0037, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0041, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0020, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0045, -0.0222, 0.0262], + ..., + [-0.0147, 0.0051, 0.0260, ..., 0.0045, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0057, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0042, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0112], + [-0.0278, -0.0054, -0.0374, ..., 0.0276, -0.0046, 0.0151], + [ 0.0089, 0.0237, -0.0076, ..., 0.0527, -0.0078, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0075, -0.0095], + [-0.0068, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0039, -0.0128, ..., 0.0060, -0.0371, -0.0046], + [ 0.0071, -0.0127, 0.0045, ..., -0.0204, 0.0039, 0.0049], + [-0.0110, -0.0195, -0.0050, ..., -0.0542, -0.0092, -0.0063], + ..., + [-0.0205, -0.0084, 0.0039, ..., -0.0199, 0.0046, 0.0039], + [-0.0220, -0.0302, -0.0101, ..., 0.0105, 0.0369, 0.0041], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[-0.0042, 0.0347, 0.0092, ..., 0.0277, -0.0208, -0.0220], + [-0.0038, 0.0043, 0.0238, ..., -0.0145, -0.0320, -0.0046], + [-0.0193, 0.0364, 0.0039, ..., -0.0077, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0065, -0.0232, -0.0464], + [-0.0110, -0.0298, -0.0195, ..., 0.0112, 0.0312, 0.0154], + [-0.0072, -0.0119, -0.0195, ..., 0.0041, 0.0109, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0078, -0.0239, -0.0123, ..., -0.0062, 0.0243, -0.0396], + [-0.0100, 0.0154, 0.0026, ..., -0.0277, -0.0027, -0.0038], + ..., + [ 0.0117, -0.0052, 0.0515, ..., 0.0327, -0.0120, -0.0074], + [ 0.0039, 0.0170, 0.0055, ..., 0.0039, 0.0280, -0.0210], + [-0.0025, 0.0072, 0.0045, ..., 0.0155, 0.0042, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [ 0.0040, 0.0081, -0.0074, ..., -0.0588, 0.0187, -0.0078], + [-0.0102, 0.0199, -0.0078, ..., 0.0080, -0.0332, -0.0076], + ..., + [-0.0052, -0.0464, 0.0277, ..., -0.0143, 0.0198, 0.0054], + [ 0.0162, 0.0011, 0.0085, ..., -0.0104, -0.0066, 0.0297], + [-0.0269, 0.0131, 0.0164, ..., 0.0084, 0.0176, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0046, ..., -0.0505, -0.0332, 0.0154], + [ 0.0208, 0.0366, 0.0007, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, 0.0058, -0.0201, ..., 0.0330, 0.0309, -0.0064], + ..., + [ 0.0028, 0.0123, 0.0153, ..., 0.0039, 0.0034, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0042, 0.0175], + [-0.0244, -0.0062, -0.0300, ..., -0.0061, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, 0.0040, 0.0075, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0050, ..., 0.0310, 0.0088, -0.0039], + [-0.0078, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0087, -0.0109, -0.0237, ..., -0.0104, 0.0157, -0.0021], + [-0.0085, 0.0205, 0.0145, ..., 0.0262, -0.0195, 0.0095], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0070]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0029, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0029, 0.0034, -0.0093], + [ 0.0017, -0.0011, 0.0398, ..., -0.0304, -0.0030, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0039], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0048, 0.0291], + [ 0.0132, -0.0004, -0.0339, ..., 0.0065, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, 0.0024], + [-0.0166, -0.0043, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0064, 0.0020, ..., 0.0240, 0.0031, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0020], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0009, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0066, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0041], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, 0.0034, 0.0021], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, -0.0042, ..., -0.0164, -0.0040, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0027, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0053, -0.0231, -0.0255, ..., -0.0068, 0.0295, -0.0046], + [-0.0255, 0.0513, 0.0030, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0035, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0064, -0.0066, -0.0364, ..., -0.0175, 0.0306, -0.0077], + [-0.0051, -0.0195, -0.0027, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0078, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0078, -0.0193, -0.0140], + [-0.0166, -0.0078, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0134, -0.0044, 0.0137, ..., 0.0021, 0.0211, -0.0275], + [ 0.0051, -0.0076, -0.0234, ..., 0.0088, -0.0084, -0.0166], + [ 0.0225, -0.0251, 0.0039, ..., 0.0072, 0.0236, 0.0077], + ..., + [ 0.0184, 0.0145, -0.0041, ..., 0.0269, -0.0085, -0.0096], + [-0.0038, 0.0044, 0.0078, ..., -0.0260, 0.0267, 0.0150], + [ 0.0102, 0.0043, 0.0212, ..., 0.0195, -0.0157, 0.0124]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0085, -0.0270, 0.0251, ..., 0.0095, 0.0052, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.0.proj.1.weight': tensor([[ 0.0018, 0.0251, 0.0957, ..., -0.0535, 0.0286, -0.0262], + [-0.0195, 0.0112, -0.1108, ..., 0.0304, -0.0312, 0.0593], + [ 0.0017, 0.1309, -0.0078, ..., -0.0058, 0.0078, -0.0282], + ..., + [ 0.0039, -0.0038, -0.0136, ..., 0.0142, -0.0222, -0.0051], + [-0.0349, -0.0645, 0.0508, ..., 0.0364, 0.0630, 0.0566], + [ 0.0192, -0.0223, 0.0010, ..., 0.0115, 0.0510, -0.0084]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_proj.weight': tensor([[ 0.0182, 0.0081, 0.0123, ..., 0.0016, -0.0015, 0.0272], + [ 0.0183, -0.0024, 0.0045, ..., 0.0009, -0.0028, 0.0209], + [-0.0061, 0.0053, -0.0056, ..., -0.0055, -0.0039, -0.0149], + ..., + [-0.0089, -0.0040, -0.0085, ..., 0.0106, -0.0037, -0.0076], + [-0.0016, -0.0033, -0.0052, ..., -0.0043, -0.0039, -0.0052], + [ 0.0081, 0.0026, 0.0023, ..., 0.0082, -0.0038, 0.0005]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_norm.weight': tensor([ 1.0742e-01, -9.8633e-02, 1.1279e-01, 3.6914e-01, -4.8056e-07, + -1.9092e-07, 2.0752e-03, 7.8870e-13, 5.4956e-15, 4.9738e-13, + -2.0961e-13, 3.3307e-16, 1.1519e-15, -4.4631e-14, 1.0868e-10, + -4.3485e-12, -3.3617e-05, 1.4827e-06, -8.6212e-04, 5.6744e-05, + 1.1505e-10, -2.5466e-11, 1.9409e-02, -4.1962e-04, 3.0518e-02, + 6.6406e-01, 1.3867e-01, -2.4414e-03, 5.6641e-02, 1.0596e-01, + 1.3279e-10, 5.4845e-14, -2.2736e-03, 1.4663e-05, 2.9755e-03, + 1.0376e-03, 6.8212e-11, -2.9650e-10, 4.8637e-05, 1.5099e-14, + 9.0639e-17, -1.2434e-14, 6.4393e-15, 4.4250e-03, 3.2471e-02, + 2.2125e-03, -3.5400e-02, -4.6875e-02, 4.8242e-01, 2.6703e-03, + -6.8665e-03, 3.4027e-03, 3.9062e-03, 1.6838e-06, 2.1210e-03, + 1.2733e-10, -1.3131e-11, 3.3396e-13, -4.5103e-16, 1.0249e-19, + -1.2967e-13, -3.8858e-15, 2.6557e-10, 3.3168e-15, 3.3203e-01, + 3.4766e-01, 4.3750e-01, 1.1768e-01, -2.4959e-07, -1.0245e-07, + 6.9580e-03, 1.5135e-12, -2.9088e-14, -3.2152e-13, 2.8866e-14, + 3.3307e-16, -1.2663e-16, -2.4425e-14, 1.8758e-11, -6.3949e-13, + 7.3910e-05, 4.0770e-05, 2.8133e-05, 5.7220e-05, -5.3478e-10, + -1.2296e-09, -8.8120e-04, -1.3580e-03, 4.8633e-01, 9.0820e-02, + 6.0938e-01, 7.9590e-02, 4.5471e-03, -2.6550e-03, 5.6389e-11, + 1.0081e-13, 2.9373e-04, 1.6117e-04, -9.7752e-06, 4.9210e-04, + -5.7116e-10, 1.4734e-10, 3.7994e-03, -3.6915e-15, 7.5287e-16, + -4.1980e-16, 2.0400e-15, 2.5269e-02, -6.6223e-03, -1.1978e-03, + 4.5117e-01, 2.8516e-01, 1.0498e-01, -7.8125e-03, 7.8125e-03, + -2.9907e-03, -4.0894e-03, 2.6375e-06, 2.1057e-03, -1.4552e-10, + 3.4333e-11, -4.5119e-13, 1.4502e-15, -3.6321e-18, 4.6896e-13, + -9.6367e-14, 4.2917e-12, 3.7303e-14], dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.self_attn.k_proj.weight': tensor([[-0.0102, 0.0103, -0.0096, ..., -0.0046, -0.0105, 0.0022], + [-0.0124, 0.0089, 0.0037, ..., 0.0029, -0.0098, 0.0126], + [-0.0073, -0.0295, 0.0043, ..., -0.0023, 0.0043, -0.0208], + ..., + [-0.0018, 0.0002, -0.0042, ..., -0.0025, -0.0005, -0.0027], + [-0.0027, -0.0030, -0.0047, ..., 0.0026, 0.0041, 0.0038], + [-0.0037, -0.0039, -0.0026, ..., 0.0029, -0.0006, 0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.k_norm.weight': tensor([ 2.9102e-01, 2.0312e-01, -2.0508e-01, 3.4961e-01, 4.7497e-07, + -2.1793e-07, 4.3335e-03, 8.1357e-13, 1.1990e-14, 5.0449e-13, + 1.8741e-13, 2.9837e-16, 1.1519e-15, -4.4853e-14, -1.0823e-10, + -4.4054e-12, 7.4863e-05, -1.2945e-07, 1.0910e-03, 5.6744e-05, + 1.2187e-10, -2.9786e-11, 1.3367e-02, 7.2861e-04, 4.1260e-02, + 6.6406e-01, 1.9336e-01, 7.6660e-02, 4.1809e-03, 2.9206e-05, + -1.3733e-10, 6.8390e-14, 2.2888e-03, 4.6492e-05, 2.9907e-03, + 1.0376e-03, 7.9126e-11, 2.9286e-10, 6.8665e-05, 1.4322e-14, + 1.1241e-15, -6.6058e-15, -6.4670e-15, 2.1118e-02, 2.9297e-02, + 2.1820e-03, 1.0254e-01, 1.1963e-01, 5.0391e-01, 7.8125e-03, + -7.8125e-03, -3.9062e-03, 4.0894e-03, -3.2783e-06, -2.0142e-03, + -6.5938e-11, -1.4353e-12, 3.4639e-13, 2.3419e-16, -4.7095e-19, + 1.4655e-13, -7.8271e-15, -2.6739e-10, 3.7470e-15, 1.7285e-01, + 3.4180e-01, 4.3164e-01, -2.6172e-01, -2.3562e-07, -1.1583e-08, + 3.5553e-03, 1.4566e-12, -2.6978e-14, 3.3751e-13, 3.1752e-14, + -3.8858e-16, 1.0278e-16, 1.5543e-14, -1.9554e-11, -6.6436e-13, + 1.1349e-04, -4.7684e-05, 1.4603e-05, 5.8413e-05, 5.3842e-10, + -1.2151e-09, 2.1744e-04, -9.0408e-04, 4.8242e-01, -8.9355e-02, + 6.0156e-01, -3.7689e-03, 5.5664e-02, 1.0547e-01, 1.1880e-11, + 2.2093e-14, -3.5095e-04, 1.6117e-04, -1.9169e-04, -4.8256e-04, + 5.7844e-10, -2.5466e-10, 3.7689e-03, 3.7748e-15, 5.4678e-15, + -8.6736e-17, 2.0817e-15, 1.4282e-02, 1.7334e-02, 1.4725e-03, + 4.5312e-01, 2.6172e-01, 1.4453e-01, 4.6082e-03, 7.6294e-03, + 1.3199e-03, 3.0975e-03, 9.3877e-07, 2.2583e-03, 1.9281e-10, + 3.6607e-11, 4.1389e-13, -1.4919e-15, 3.5508e-18, -4.6896e-13, + -9.6811e-14, 1.1084e-12, 3.7303e-14], dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.self_attn.v_proj.weight': tensor([[-0.0022, 0.0098, 0.0055, ..., -0.0032, -0.0040, -0.0063], + [ 0.0002, 0.0038, -0.0006, ..., -0.0014, 0.0046, 0.0002], + [ 0.0142, 0.0043, 0.0038, ..., -0.0041, 0.0016, 0.0063], + ..., + [-0.0006, 0.0021, -0.0019, ..., 0.0025, -0.0025, -0.0029], + [ 0.0063, -0.0032, 0.0037, ..., 0.0059, -0.0058, 0.0040], + [-0.0090, 0.0049, -0.0024, ..., 0.0055, -0.0038, -0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.output_proj.weight': tensor([[ 0.0145, -0.0216, -0.0175, ..., 0.0042, 0.0087, -0.0032], + [ 0.0022, -0.0068, 0.0124, ..., -0.0063, -0.0024, -0.0059], + [ 0.0040, 0.0022, 0.0052, ..., 0.0009, -0.0058, 0.0017], + ..., + [ 0.0055, -0.0001, 0.0048, ..., 0.0042, -0.0014, 0.0009], + [-0.0055, -0.0045, 0.0275, ..., 0.0036, -0.0033, 0.0161], + [-0.0084, 0.0095, -0.0052, ..., 0.0051, 0.0051, -0.0053]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_proj.weight': tensor([[-0.0034, -0.0093, -0.0017, ..., -0.0048, -0.0034, -0.0014], + [-0.0096, 0.0098, 0.0019, ..., 0.0072, 0.0025, -0.0030], + [-0.0023, -0.0084, 0.0025, ..., -0.0004, 0.0062, 0.0029], + ..., + [ 0.0030, 0.0071, 0.0009, ..., 0.0047, -0.0041, 0.0082], + [-0.0015, -0.0042, -0.0017, ..., -0.0048, 0.0043, -0.0067], + [-0.0060, 0.0023, 0.0009, ..., -0.0008, 0.0033, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_norm.weight': tensor([ 6.0120e-03, 8.8811e-06, 2.7740e-11, 1.4526e-02, -9.6436e-03, + 1.5442e-02, 2.2054e-06, 5.4550e-04, 3.4912e-02, 2.6978e-02, + 1.6880e-08, 2.8076e-02, 5.7812e-01, 9.3994e-03, 1.0437e-02, + 4.0894e-03, 4.6680e-01, 1.1086e-05, 1.1215e-03, 1.0803e-02, + 8.8501e-03, 7.3547e-03, 8.3496e-02, 1.8120e-05, 1.8433e-02, + 1.0681e-03, 7.9346e-03, -8.4400e-05, 1.2436e-03, 1.6708e-03, + 3.5156e-01, 9.0942e-03, 5.0306e-05, 4.6680e-01, 5.4297e-01, + 1.3828e-04, -5.1022e-05, 3.1494e-02, 1.5137e-01, 2.5635e-02, + 2.0996e-02, -4.6790e-06, 1.3642e-10, 4.1809e-03, 9.4223e-04, + 3.3203e-02, 1.2024e-02, 1.0605e-03, 4.0894e-03, 8.7891e-03, + 1.8539e-03, -7.4863e-05, -4.2915e-05, 1.5106e-03, -3.8300e-03, + 1.6689e-04, 1.9653e-02, 1.1396e-04, 1.7090e-02, 1.1475e-02, + 1.9932e-04, -7.8082e-06, 1.2329e-02, 7.2632e-03, 4.0293e-05, + 1.3550e-02, 9.9945e-04, 2.5000e-01, 8.8501e-04, 2.1851e-02, + 4.6875e-01, -1.2875e-05, 4.5410e-02, 8.4229e-03, 1.7695e-08, + 7.0190e-03, 1.6632e-03, 7.0801e-03, -1.3590e-05, 2.0752e-02, + 2.3438e-02, 4.2725e-03, 1.7188e-01, 7.6294e-03, 1.5869e-03, + 2.4567e-03, 1.3000e-02, -1.3530e-05, 1.0395e-04, 5.0781e-01, + 2.6550e-03, 8.9645e-05, 1.1520e-03, 3.8147e-06, 9.3384e-03, + 4.5117e-01, 4.3640e-03, 1.0538e-04, -3.7126e-13, 9.4223e-04, + 3.9062e-01, 1.3351e-05, -9.0332e-03, 1.7822e-02, 2.3071e-02, + 9.3460e-05, -1.5080e-05, 1.3769e-05, 1.6708e-03, 1.7944e-02, + 1.1597e-02, 8.5449e-04, 4.2419e-03, 1.2329e-02, -3.6316e-03, + 1.0395e-04, 1.1230e-02, -2.8534e-03, -1.1292e-03, 1.8311e-03, + 1.0300e-04, 6.3782e-03, 1.1368e-03, -2.2650e-05, 1.8978e-04, + 1.1414e-02, -7.1526e-06, 1.0133e-05], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_proj.weight': tensor([[ 1.5411e-03, 7.2861e-04, 1.9150e-03, ..., -6.7520e-04, + -2.4605e-04, 6.1035e-04], + [-3.6926e-03, -2.0599e-03, -3.9368e-03, ..., -2.6093e-03, + -3.9673e-03, -3.0060e-03], + [-1.7700e-03, -1.4877e-03, -2.4872e-03, ..., -2.8381e-03, + -5.6458e-03, -9.6130e-04], + ..., + [ 2.5635e-03, -2.4719e-03, 1.8768e-03, ..., 5.2643e-04, + 2.3041e-03, -5.7220e-04], + [-8.8882e-04, 5.3024e-04, -2.3193e-03, ..., -3.0212e-03, + 5.0354e-04, 2.3193e-03], + [ 2.9297e-03, 1.8997e-03, -3.2501e-03, ..., -4.2725e-03, + 6.5613e-04, 5.7697e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_norm.weight': tensor([ 5.9814e-03, -5.5730e-06, -2.7740e-11, 1.4343e-02, 9.7046e-03, + 1.5442e-02, 2.1756e-06, -5.4550e-04, 3.3936e-02, 2.7100e-02, + -5.9954e-09, 2.8320e-02, 5.8203e-01, 9.3384e-03, 9.8877e-03, + 4.0588e-03, 4.6680e-01, 1.1325e-05, 9.9182e-04, 1.0620e-02, + 8.7280e-03, 7.3853e-03, 8.3496e-02, 3.7670e-05, 1.8311e-02, + 1.0757e-03, 7.5073e-03, -8.3923e-05, 6.2561e-04, 1.6403e-03, + 3.5156e-01, 8.9111e-03, 5.0068e-05, 4.6680e-01, 5.4297e-01, + -2.0981e-04, 5.8413e-05, 3.1982e-02, 1.5137e-01, 2.5879e-02, + 2.0996e-02, 9.7275e-05, 1.9936e-09, 4.0283e-03, -2.2793e-04, + 3.2715e-02, 1.1902e-02, 1.0757e-03, 4.0283e-03, 8.7280e-03, + 1.8463e-03, -7.4863e-05, 4.2915e-05, 1.5030e-03, 3.9978e-03, + 2.1076e-04, 1.9531e-02, 1.3351e-04, 1.7334e-02, 1.1536e-02, + 1.9932e-04, -8.1062e-06, 1.2085e-02, 7.2632e-03, 4.0054e-05, + 1.3733e-02, 9.9182e-04, 2.5000e-01, 8.9645e-04, 2.1729e-02, + 4.6875e-01, 1.2696e-05, 4.3945e-02, 8.4229e-03, -1.9674e-08, + 7.1411e-03, 5.5313e-04, 7.0190e-03, -1.4782e-05, 2.0752e-02, + 2.3560e-02, 4.2419e-03, 1.7188e-01, 7.4463e-03, 1.6251e-03, + 2.4872e-03, 1.4038e-02, 1.3590e-05, 1.0490e-04, 5.0781e-01, + 2.6398e-03, -1.2398e-04, 7.5912e-04, 3.7253e-06, 9.3384e-03, + 4.5117e-01, -4.2419e-03, -9.1553e-05, 3.7126e-13, 9.1934e-04, + 3.9062e-01, 1.3351e-05, 1.0193e-02, 1.7822e-02, 2.3071e-02, + 9.2983e-05, 1.7524e-05, 1.3888e-05, 8.1253e-04, 1.7944e-02, + 1.1841e-02, 8.5449e-04, 4.2114e-03, 1.2268e-02, 3.6163e-03, + -1.6785e-04, 1.1292e-02, 2.8687e-03, 1.3123e-03, 1.8845e-03, + 9.7275e-05, 6.5918e-03, 1.1139e-03, -9.0599e-05, 1.0633e-04, + 1.1292e-02, -6.8247e-06, -3.7700e-06], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.v_proj.weight': tensor([[-9.8877e-03, 2.1820e-03, 1.2695e-02, ..., 2.8534e-03, + 2.6550e-03, -1.0193e-02], + [-2.2461e-02, 8.1635e-04, 7.7209e-03, ..., -3.2043e-03, + -6.1035e-03, 9.3937e-05], + [-1.2573e-02, 2.4567e-03, 2.9297e-03, ..., -2.2888e-03, + 1.3000e-02, -1.2268e-02], + ..., + [ 1.0925e-02, 1.8692e-03, 1.6861e-03, ..., 2.5940e-04, + 2.9755e-03, 1.8799e-02], + [-9.5215e-03, 2.5330e-03, 1.0315e-02, ..., 1.8463e-03, + -1.7700e-02, -3.7079e-03], + [-1.6479e-02, -2.8839e-03, 7.1716e-03, ..., -2.0599e-03, + -6.0425e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.output_proj.weight': tensor([[ 0.0064, -0.0018, 0.0056, ..., -0.0030, 0.0018, 0.0080], + [ 0.0080, 0.0066, -0.0012, ..., 0.0162, 0.0041, -0.0110], + [ 0.0052, -0.0030, -0.0008, ..., -0.0004, -0.0073, -0.0052], + ..., + [-0.0071, -0.0077, -0.0051, ..., -0.0079, -0.0047, -0.0085], + [ 0.0101, -0.0025, 0.0059, ..., -0.0142, -0.0183, 0.0055], + [ 0.0106, -0.0081, -0.0041, ..., -0.0070, -0.0037, -0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_img.weight': tensor([[ 0.0067, 0.0119, -0.0041, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0075, 0.0249, 0.0291], + [-0.0095, 0.0046, -0.0115, ..., -0.0342, -0.0327, 0.0061], + ..., + [ 0.0074, -0.0067, 0.0040, ..., 0.0082, 0.0286, -0.0165], + [ 0.0078, 0.0013, -0.0102, ..., -0.0100, 0.0142, -0.0309], + [ 0.0242, -0.0106, 0.0255, ..., -0.0227, -0.0043, -0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, -0.0010], + [ 0.0026, 0.0120, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0085, ..., -0.0057, -0.0046, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0039, 0.0054], + [-0.0160, 0.0055, -0.0036, ..., 0.0133, -0.0173, -0.0049], + [-0.0051, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img.weight': tensor([[-0.0510, -0.0021, -0.0364, ..., -0.0225, -0.0143, -0.0055], + [-0.0157, 0.0053, -0.0364, ..., 0.0099, 0.0117, -0.0036], + [ 0.0025, 0.0615, 0.0056, ..., -0.0205, -0.0306, -0.0114], + ..., + [ 0.0008, -0.0018, -0.0103, ..., -0.0211, -0.0012, 0.0288], + [ 0.0266, -0.0115, -0.0056, ..., -0.0420, -0.0282, -0.0008], + [ 0.0286, 0.0325, 0.0095, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.mlp.layer1.weight': tensor([[ 0.0082, 0.0159, 0.0067, ..., 0.0008, 0.0096, 0.0064], + [ 0.0050, -0.0152, -0.0035, ..., -0.0021, -0.0115, 0.0064], + [ 0.0062, -0.0057, -0.0039, ..., 0.0176, -0.0176, 0.0121], + ..., + [-0.0037, 0.0004, 0.0240, ..., 0.0054, 0.0075, 0.0029], + [-0.0030, 0.0186, 0.0068, ..., 0.0021, -0.0079, 0.0085], + [ 0.0050, 0.0027, -0.0013, ..., 0.0029, -0.0074, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.mlp.layer2.weight': tensor([[-0.0077, -0.0042, -0.0074, ..., -0.0038, -0.0261, -0.0078], + [ 0.0042, 0.0085, -0.0029, ..., -0.0162, -0.0216, 0.0073], + [-0.0062, 0.0092, 0.0090, ..., 0.0012, -0.0023, 0.0053], + ..., + [-0.0044, -0.0049, -0.0091, ..., -0.0069, -0.0026, -0.0082], + [ 0.0079, 0.0125, 0.0215, ..., 0.0008, 0.0056, 0.0114], + [ 0.0085, 0.0122, 0.0012, ..., 0.0050, -0.0021, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0610, -0.0593, -0.0579, ..., 0.0030, 0.0031, 0.0031], + [ 0.0591, 0.0583, 0.0562, ..., -0.0024, -0.0025, -0.0026], + [ 0.0583, 0.0576, 0.0564, ..., -0.0026, -0.0027, -0.0027], + ..., + [ 0.0630, 0.0615, 0.0598, ..., -0.0024, -0.0025, -0.0025], + [ 0.0598, 0.0583, 0.0566, ..., -0.0022, -0.0023, -0.0022], + [ 0.0598, 0.0583, 0.0571, ..., -0.0030, -0.0031, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0009, 0.0014, 0.0015, ..., 0.0050, 0.0036, 0.0003], + [-0.0025, -0.0015, -0.0016, ..., 0.0022, 0.0022, 0.0024], + [-0.0035, 0.0033, 0.0031, ..., 0.0059, 0.0060, 0.0034], + ..., + [-0.0432, 0.0500, 0.0493, ..., 0.0505, 0.0464, 0.0437], + [-0.0058, 0.0069, 0.0061, ..., 0.0053, 0.0044, 0.0057], + [ 0.0096, -0.0104, -0.0103, ..., -0.0099, -0.0096, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_cross_attn.1.weight': tensor([[-8.5449e-03, 4.4556e-03, 4.4861e-03, ..., -6.7520e-04, + -6.0654e-04, -7.7438e-04], + [-3.8818e-02, -3.7109e-02, -3.7109e-02, ..., 2.5177e-04, + 2.5940e-04, 2.6512e-04], + [ 4.1748e-02, 4.0283e-02, 4.0039e-02, ..., -7.0095e-05, + -5.1022e-05, -7.5340e-05], + ..., + [ 3.9062e-02, 3.8086e-02, 3.7109e-02, ..., -3.3569e-04, + -2.6703e-04, -3.0327e-04], + [ 4.9072e-02, 4.8096e-02, 4.6631e-02, ..., 5.0354e-04, + 6.1417e-04, 5.4550e-04], + [ 3.8818e-02, 3.7598e-02, 3.6621e-02, ..., -5.6624e-06, + 7.0572e-05, 2.1935e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0129, -0.0072, 0.0067, ..., 0.0072, 0.0072, 0.0068], + [ 0.0052, 0.0034, -0.0049, ..., -0.0034, -0.0034, -0.0048], + [-0.0037, -0.0036, 0.0035, ..., 0.0036, 0.0035, 0.0036], + ..., + [-0.0063, -0.0205, 0.0364, ..., 0.0203, 0.0432, 0.0320], + [ 0.0103, 0.0139, 0.0101, ..., -0.0131, 0.0050, 0.0099], + [-0.0137, -0.0046, -0.0113, ..., 0.0045, -0.0094, -0.0110]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_mlp.1.weight': tensor([[-0.0049, -0.0049, 0.0038, ..., 0.0004, 0.0007, 0.0006], + [ 0.0640, 0.0618, 0.0605, ..., -0.0024, -0.0027, -0.0027], + [ 0.0554, 0.0542, 0.0525, ..., -0.0037, -0.0040, -0.0040], + ..., + [ 0.0952, 0.0928, 0.0908, ..., -0.0008, -0.0009, -0.0010], + [-0.0038, -0.0038, -0.0038, ..., -0.0004, -0.0005, -0.0004], + [ 0.0747, 0.0723, 0.0708, ..., -0.0012, -0.0012, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_mlp.2.weight': tensor([[-0.0077, 0.0091, 0.0107, ..., 0.0072, 0.0082, 0.0060], + [-0.0007, 0.0023, 0.0033, ..., 0.0031, 0.0007, 0.0009], + [-0.0017, 0.0060, 0.0061, ..., 0.0044, 0.0015, 0.0057], + ..., + [ 0.0036, 0.0747, 0.1025, ..., 0.0635, -0.0044, 0.0496], + [ 0.0114, 0.0165, 0.0096, ..., 0.0280, -0.0114, 0.0233], + [-0.0089, 0.0033, -0.0003, ..., 0.0041, 0.0087, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.before_proj.weight': tensor([[ 0.0028, 0.0002, -0.0006, ..., 0.0062, 0.0011, -0.0013], + [-0.0094, -0.0055, -0.0041, ..., -0.0078, -0.0024, 0.0052], + [-0.0075, -0.0059, -0.0033, ..., -0.0015, -0.0036, -0.0027], + ..., + [-0.0011, 0.0017, -0.0047, ..., 0.0033, -0.0049, -0.0004], + [-0.0037, -0.0017, 0.0011, ..., -0.0020, -0.0052, -0.0042], + [ 0.0027, 0.0028, -0.0021, ..., 0.0045, 0.0005, -0.0034]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.before_proj.bias': tensor([-0.0199, 0.0027, 0.0018, ..., 0.0016, -0.0040, -0.0239], + dtype=torch.bfloat16), 'net.control_blocks_0.0.after_proj.weight': tensor([[-0.0188, -0.0083, 0.0017, ..., -0.0049, -0.0119, 0.0113], + [ 0.0189, -0.0134, 0.0063, ..., -0.0007, 0.0084, -0.0087], + [ 0.0009, -0.0009, -0.0043, ..., -0.0043, 0.0015, -0.0023], + ..., + [ 0.0029, -0.0039, 0.0083, ..., -0.0016, -0.0035, 0.0015], + [ 0.0276, -0.0150, 0.0077, ..., -0.0062, -0.0171, -0.0134], + [-0.0036, -0.0123, 0.0019, ..., 0.0026, -0.0048, 0.0112]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.after_proj.bias': tensor([ 0.0075, -0.0165, -0.0026, ..., -0.0014, -0.0140, 0.0598], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_proj.weight': tensor([[-0.0220, 0.0138, -0.0125, ..., -0.0093, -0.0069, 0.0132], + [ 0.0049, 0.0032, -0.0035, ..., 0.0193, -0.0050, -0.0388], + [ 0.0012, 0.0103, 0.0013, ..., -0.0023, 0.0043, -0.0078], + ..., + [ 0.0009, 0.0050, -0.0006, ..., 0.0024, -0.0010, -0.0002], + [ 0.0021, -0.0055, -0.0013, ..., 0.0013, -0.0062, 0.0017], + [ 0.0025, -0.0006, 0.0052, ..., 0.0031, -0.0019, -0.0037]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_norm.weight': tensor([ 3.9062e-01, 1.2578e+00, 5.6396e-02, 1.1953e+00, 8.9844e-01, + 4.6945e-34, -1.9184e-34, -1.5573e-34, -6.8310e-34, 3.3087e-22, + 8.9676e-34, -4.9653e-34, -9.3287e-35, -3.7014e-34, -7.2523e-34, + -1.0653e-33, 1.4551e-01, 3.8338e-04, -2.1696e-05, 2.2339e-02, + 2.4319e-04, -5.7697e-05, 4.6289e-01, 8.8672e-01, 1.3516e+00, + 4.4531e-01, -8.6426e-02, 6.0547e-01, 8.4229e-03, -2.5022e-30, + 1.9259e-34, 1.6852e-33, -2.3127e-05, 8.9264e-04, 8.7738e-05, + -9.7656e-03, 3.1281e-04, -1.0909e-34, -4.9277e-35, 1.1977e-33, + -1.3542e-33, 2.1507e-22, 6.3496e-34, 1.0859e+00, 3.2422e-01, + 2.1973e-01, 3.7500e-01, 1.3438e+00, 2.0630e-02, -4.4556e-03, + -6.2256e-03, 2.3926e-01, 1.7395e-03, -4.6042e-34, -3.7917e-34, + -3.0333e-33, 7.4158e-03, -6.0100e-25, 1.9019e-33, 2.0613e-34, + 3.8556e-35, 4.0813e-35, -1.6325e-34, -2.7836e-35, 5.0391e-01, + 4.1406e-01, 6.3281e-01, 4.9805e-01, 1.6797e-01, 5.2662e-34, + -1.2883e-35, 4.4838e-34, 7.4028e-34, -3.2467e-23, 2.1667e-34, + -1.5528e-33, 2.7836e-35, -1.0833e-34, 9.3287e-34, 8.1852e-34, + 4.6387e-03, 8.9264e-04, -4.4346e-05, 4.4250e-04, 2.3079e-04, + 3.5889e-02, 1.1953e+00, 2.5391e-01, 4.5117e-01, 1.3359e+00, + 2.7734e-01, 2.4414e-01, 1.9455e-03, -2.0338e-30, 2.1215e-34, + -1.3120e-33, -3.0279e-05, 3.4714e-04, -1.6212e-05, 1.3550e-02, + 3.8719e-04, -1.1887e-34, -4.1001e-35, -1.6777e-34, -1.1857e-33, + -4.1359e-24, 4.2130e-33, 2.8711e-01, 9.4141e-01, 1.2266e+00, + 1.4844e+00, 1.9336e-01, 8.3594e-01, -3.9062e-03, -8.2397e-04, + -4.7913e-03, 2.5558e-04, 1.3181e-33, -3.4005e-34, 2.1787e-33, + 6.7520e-04, -7.9616e-24, 2.3232e-33, 1.7379e-34, -1.0909e-35, + 1.1134e-34, -6.9213e-35, 6.3947e-35], dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.self_attn.k_proj.weight': tensor([[ 0.0016, -0.0195, 0.0078, ..., 0.0172, -0.0051, -0.0264], + [ 0.0066, -0.0129, -0.0035, ..., -0.0199, 0.0034, -0.0095], + [ 0.0092, 0.0063, -0.0018, ..., 0.0054, -0.0044, 0.0065], + ..., + [-0.0085, 0.0008, -0.0060, ..., 0.0082, 0.0109, -0.0054], + [-0.0095, -0.0052, 0.0011, ..., -0.0036, 0.0062, -0.0051], + [-0.0058, -0.0028, -0.0025, ..., -0.0032, 0.0117, -0.0054]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.k_norm.weight': tensor([ 5.0391e-01, 1.2812e+00, 2.3438e-01, 1.2109e+00, 8.9844e-01, + 4.5139e-34, 1.3918e-34, 1.1586e-34, 1.5167e-33, 3.3087e-22, + -2.8137e-34, -2.0162e-34, 4.4387e-35, 1.9410e-34, -6.7708e-34, + 6.0938e-35, 1.4551e-01, 3.6812e-04, 1.2875e-04, 5.5420e-02, + 2.4319e-04, 1.2493e-04, 6.8750e-01, 8.9062e-01, 1.3516e+00, + 4.5117e-01, 2.1289e-01, 6.4453e-01, -8.8501e-03, 1.6147e-30, + 1.0382e-34, -7.8993e-35, 1.4782e-04, 8.6975e-04, 1.5545e-04, + 4.5410e-02, 3.0708e-04, -5.0781e-35, -2.4262e-35, 4.1829e-34, + -3.9120e-34, -8.5200e-23, 4.7245e-34, 1.0781e+00, 3.8672e-01, + -6.3477e-02, -3.8672e-01, 1.3359e+00, 1.5527e-01, 2.4567e-03, + -6.1646e-03, 2.3340e-01, 1.7395e-03, -2.5278e-34, 1.2940e-34, + 1.1857e-33, 7.1411e-03, -6.0746e-25, -3.3102e-34, 1.3015e-34, + -1.2413e-35, -6.1314e-35, 1.6852e-34, -2.2569e-35, 5.1562e-01, + 5.5078e-01, 6.2891e-01, 4.3359e-01, -2.4414e-01, -6.9514e-34, + 1.1360e-34, 2.0162e-34, -3.6352e-33, 1.5406e-23, -2.9190e-34, + -4.9051e-34, 1.9184e-35, 8.0874e-35, -2.1366e-34, 5.5972e-34, + 5.1575e-03, 8.8501e-04, 1.3161e-04, 4.2725e-04, 2.3460e-04, + 3.5889e-02, 1.1875e+00, 3.1836e-01, 4.7070e-01, 1.3438e+00, + 2.7539e-01, 3.0273e-01, 8.3923e-04, -1.8612e-30, 1.5799e-34, + -3.1296e-33, 1.0633e-04, 4.0627e-04, 4.3869e-05, 4.4678e-02, + 3.8528e-04, 3.6864e-35, -4.5515e-35, -1.3993e-34, -1.0472e-33, + 3.0632e-24, 2.0222e-33, 4.6094e-01, 9.4531e-01, 1.2344e+00, + 1.4844e+00, 2.3438e-01, 8.4375e-01, -6.0425e-03, -7.6294e-05, + -2.0703e-01, 3.1090e-04, -6.5000e-34, -3.7014e-34, -1.4264e-33, + 1.1587e-04, 3.5931e-24, 4.6945e-34, -1.0006e-34, -3.7616e-35, + -1.0382e-34, 3.2162e-35, -4.7396e-35], dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.self_attn.v_proj.weight': tensor([[ 0.0028, -0.0034, 0.0063, ..., -0.0203, -0.0049, 0.0070], + [ 0.0029, -0.0100, 0.0068, ..., 0.0069, -0.0041, -0.0063], + [ 0.0052, -0.0027, -0.0053, ..., -0.0231, -0.0046, -0.0111], + ..., + [ 0.0065, 0.0164, 0.0194, ..., -0.0010, 0.0004, -0.0175], + [-0.0047, -0.0121, 0.0077, ..., 0.0072, -0.0036, 0.0110], + [ 0.0112, -0.0040, -0.0160, ..., 0.0114, 0.0090, 0.0106]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.output_proj.weight': tensor([[-0.0054, -0.0135, -0.0113, ..., 0.0189, -0.0115, 0.0073], + [ 0.0178, -0.0056, -0.0044, ..., -0.0254, -0.0120, -0.0076], + [-0.0027, 0.0036, 0.0038, ..., -0.0048, 0.0053, 0.0042], + ..., + [ 0.0027, -0.0137, -0.0098, ..., -0.0210, -0.0160, -0.0188], + [ 0.0042, -0.0032, -0.0053, ..., -0.0211, 0.0063, -0.0063], + [ 0.0143, 0.0056, -0.0114, ..., -0.0033, 0.0105, -0.0081]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_proj.weight': tensor([[-0.0164, 0.0239, 0.0057, ..., 0.0036, -0.0100, 0.0140], + [ 0.0244, -0.0101, -0.0052, ..., -0.0085, 0.0110, -0.0208], + [ 0.0104, 0.0075, -0.0023, ..., -0.0057, 0.0073, -0.0005], + ..., + [ 0.0092, -0.0036, 0.0033, ..., 0.0065, -0.0099, 0.0070], + [-0.0061, -0.0043, -0.0024, ..., -0.0044, -0.0095, 0.0100], + [ 0.0219, 0.0141, -0.0095, ..., -0.0045, 0.0086, -0.0172]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_norm.weight': tensor([ 2.3682e-02, 4.8047e-01, 1.7822e-02, 8.7891e-03, 6.7749e-03, + 2.0606e-12, 9.5367e-06, 5.5469e-01, -1.8332e-12, 4.4727e-01, + 1.8799e-02, 1.9287e-02, 7.0496e-03, 2.2095e-02, 1.7944e-02, + 1.7456e-02, 6.0425e-03, 1.8215e-04, -3.6001e-05, 2.9297e-02, + 2.6562e-01, 2.2461e-02, 6.5231e-04, 1.8433e-02, 1.0757e-03, + 2.6131e-04, -1.7548e-04, 3.0884e-02, 6.5327e-05, 4.2969e-01, + -1.1969e-04, -4.2915e-05, 9.6436e-03, 1.4114e-04, 7.6904e-03, + 1.0376e-02, 2.0146e-05, 4.5703e-01, 1.5259e-02, 2.7771e-03, + 3.6377e-02, 1.8921e-02, 5.9204e-03, 2.8133e-05, 5.4321e-03, + 7.6294e-03, 1.8616e-03, 1.7578e-02, 2.8076e-02, 1.4893e-02, + 4.9438e-03, 4.4336e-01, 1.0071e-02, 1.5869e-02, 3.1494e-02, + 1.8692e-04, 6.8359e-02, 1.6479e-02, 1.5869e-02, 2.5391e-02, + 4.0770e-05, 1.1475e-02, 5.7373e-03, -4.8876e-05, 1.2756e-02, + 2.5482e-03, -1.8597e-05, -2.7776e-05, 8.9722e-03, 4.7852e-02, + -5.2387e-08, 6.1951e-03, 4.6143e-02, 3.2471e-02, 4.3640e-03, + 8.3008e-03, 2.9755e-03, 1.7578e-02, 2.1973e-02, 1.4648e-02, + -2.2793e-04, 1.1963e-02, 3.2471e-02, -2.8253e-05, 2.9419e-02, + 1.2360e-03, 4.3030e-03, 3.0518e-02, 1.1658e-02, 5.5847e-03, + 1.3306e-02, 5.9891e-04, 1.1292e-02, 2.1729e-02, 1.3306e-02, + -1.0395e-04, 1.9287e-02, 3.3203e-02, 1.5991e-02, 4.1748e-02, + 2.5513e-02, 2.3438e-02, 1.6357e-02, 1.0443e-04, -2.0373e-10, + 1.3733e-02, 4.6158e-04, 2.2461e-02, 1.6479e-02, 1.9531e-02, + 1.3184e-02, 5.8594e-03, 2.1118e-02, 5.0781e-01, 1.1063e-03, + 1.1353e-02, 7.8735e-03, 1.2512e-02, 5.1575e-03, 5.2002e-02, + 2.3499e-03, 2.3651e-03, -1.4210e-04, 1.2451e-02, 6.4087e-03, + 1.3367e-02, 2.0264e-02, 2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_proj.weight': tensor([[-0.0010, -0.0033, 0.0038, ..., -0.0006, 0.0034, 0.0026], + [ 0.0116, -0.0035, -0.0040, ..., 0.0050, 0.0197, 0.0012], + [ 0.0023, 0.0036, 0.0016, ..., -0.0006, -0.0019, 0.0029], + ..., + [-0.0017, -0.0049, -0.0036, ..., -0.0004, -0.0015, 0.0031], + [-0.0016, -0.0053, -0.0125, ..., 0.0029, 0.0045, 0.0010], + [ 0.0078, -0.0004, -0.0020, ..., -0.0027, 0.0015, -0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_norm.weight': tensor([ 2.4658e-02, 4.8047e-01, 1.7578e-02, 8.8501e-03, 6.7749e-03, + 2.1885e-12, 9.4771e-06, 5.5469e-01, 1.8474e-12, 4.4727e-01, + 1.9409e-02, 1.9165e-02, 7.0190e-03, 2.1973e-02, 1.7822e-02, + 1.7334e-02, 6.2866e-03, 1.7929e-04, -3.7432e-05, 2.9175e-02, + 2.6562e-01, 2.1973e-02, 6.5613e-04, 1.8555e-02, 1.0986e-03, + 2.6321e-04, -1.7357e-04, 3.0640e-02, 6.5327e-05, 4.2969e-01, + -1.1921e-04, -4.4823e-05, 9.4604e-03, 1.4019e-04, 7.6904e-03, + 1.0315e-02, 2.0504e-05, 4.5703e-01, 1.5259e-02, 2.6550e-03, + 3.5400e-02, 1.8921e-02, 5.5847e-03, 2.5511e-05, 5.3406e-03, + 7.6904e-03, 1.8768e-03, 1.7334e-02, 2.8198e-02, 1.5015e-02, + 4.8828e-03, 4.4336e-01, 1.0132e-02, 1.5747e-02, 3.1250e-02, + 1.8597e-04, 6.8848e-02, 1.6235e-02, 1.5869e-02, 2.5513e-02, + 4.1485e-05, 1.1536e-02, 5.7373e-03, -5.0783e-05, 1.2756e-02, + 2.4872e-03, -1.9431e-05, -2.7418e-05, 9.0942e-03, 4.7852e-02, + -1.1316e-07, 6.3477e-03, 4.6631e-02, 3.2471e-02, 4.4250e-03, + 8.6670e-03, 2.9297e-03, 1.7090e-02, 2.1973e-02, 1.4771e-02, + -2.2316e-04, 1.1780e-02, 3.1982e-02, -2.7776e-05, 2.9663e-02, + 1.0910e-03, 4.2725e-03, 3.0273e-02, 1.1963e-02, 5.7373e-03, + 1.3794e-02, 6.2561e-04, 1.1292e-02, 2.1606e-02, 1.3123e-02, + -1.0586e-04, 1.8677e-02, 3.2959e-02, 1.5869e-02, 4.1016e-02, + 2.6367e-02, 2.3560e-02, 1.6602e-02, 1.0347e-04, 2.0373e-10, + 1.3672e-02, 4.5013e-04, 2.2217e-02, 1.6602e-02, 1.9287e-02, + 1.3184e-02, 5.5847e-03, 2.0996e-02, 5.0781e-01, 1.0986e-03, + 1.1292e-02, 7.8735e-03, 1.2329e-02, 5.3711e-03, 5.2002e-02, + 2.3956e-03, 2.3956e-03, -1.4210e-04, 1.2634e-02, 6.3477e-03, + 1.3245e-02, 2.0264e-02, 2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.v_proj.weight': tensor([[-2.2583e-02, 2.2736e-03, 2.2430e-03, ..., -8.6784e-05, + -1.5442e-02, -2.7832e-02], + [-4.1199e-03, 1.0803e-02, 1.0376e-02, ..., 7.7820e-03, + 6.6833e-03, 7.7209e-03], + [-3.9307e-02, -5.0049e-03, -7.4158e-03, ..., -9.3994e-03, + -2.7710e-02, -1.2451e-02], + ..., + [-4.1748e-02, 7.0190e-04, -1.5625e-02, ..., 1.8158e-03, + 2.1729e-02, 9.3994e-03], + [ 6.5308e-03, 6.1646e-03, -4.4556e-03, ..., 3.6469e-03, + 8.6670e-03, 8.1787e-03], + [ 1.6235e-02, 4.6692e-03, -3.8757e-03, ..., 2.0294e-03, + 1.1353e-02, -4.6015e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.output_proj.weight': tensor([[ 0.0071, 0.0040, 0.0148, ..., 0.0079, 0.0043, -0.0017], + [-0.0080, -0.0045, -0.0041, ..., -0.0087, 0.0038, 0.0112], + [-0.0005, 0.0038, -0.0038, ..., 0.0015, -0.0027, -0.0022], + ..., + [ 0.0054, 0.0052, 0.0035, ..., -0.0064, -0.0051, 0.0078], + [ 0.0128, 0.0302, -0.0167, ..., -0.0222, 0.0045, 0.0425], + [-0.0012, -0.0087, -0.0229, ..., -0.0104, 0.0020, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_img.weight': tensor([[ 0.0079, 0.0020, 0.0430, ..., -0.0327, 0.0054, 0.0148], + [-0.0278, 0.0280, 0.0249, ..., -0.0068, -0.0378, 0.0271], + [ 0.0139, -0.0116, 0.0312, ..., -0.0075, 0.0253, -0.0110], + ..., + [ 0.0245, -0.0150, 0.0033, ..., -0.0452, 0.0216, -0.0209], + [-0.0081, 0.0073, -0.0471, ..., -0.0109, 0.0157, -0.0088], + [ 0.0078, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0038, -0.0123, -0.0115], + [ 0.0135, 0.0061, -0.0188, ..., 0.0157, 0.0082, -0.0042], + [-0.0306, -0.0028, -0.0272, ..., 0.0101, 0.0337, 0.0039], + ..., + [-0.0040, -0.0040, 0.0047, ..., 0.0209, -0.0103, 0.0073], + [ 0.0193, 0.0210, 0.0123, ..., -0.0145, 0.0117, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0110, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img.weight': tensor([[ 0.0058, -0.0056, 0.0173, ..., 0.0083, 0.0356, 0.0217], + [-0.0498, -0.0051, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0108, -0.0044, -0.0139, ..., 0.0039, 0.0269, 0.0077], + ..., + [-0.0043, 0.0125, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0077, -0.0162, ..., -0.0054, -0.0075, -0.0121], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0138, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.mlp.layer1.weight': tensor([[-0.0176, 0.0156, -0.0038, ..., -0.0092, -0.0087, -0.0221], + [-0.0026, -0.0106, 0.0096, ..., -0.0039, -0.0060, -0.0166], + [-0.0065, 0.0032, 0.0022, ..., -0.0078, -0.0032, -0.0050], + ..., + [ 0.0251, 0.0005, -0.0049, ..., 0.0010, 0.0085, 0.0071], + [-0.0251, 0.0047, -0.0145, ..., 0.0160, -0.0053, 0.0062], + [ 0.0092, -0.0067, -0.0008, ..., -0.0135, 0.0120, -0.0248]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.mlp.layer2.weight': tensor([[-0.0166, -0.0135, 0.0115, ..., -0.0025, 0.0060, -0.0061], + [-0.0049, 0.0071, 0.0084, ..., -0.0048, 0.0110, 0.0041], + [-0.0077, -0.0070, 0.0029, ..., -0.0098, 0.0082, -0.0077], + ..., + [-0.0027, 0.0090, 0.0045, ..., 0.0143, -0.0068, 0.0093], + [ 0.0049, 0.0103, 0.0046, ..., -0.0059, -0.0005, 0.0159], + [ 0.0105, -0.0038, -0.0099, ..., 0.0050, -0.0029, -0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0447, -0.0435, -0.0430, ..., 0.0050, 0.0053, 0.0053], + [ 0.0408, 0.0396, 0.0381, ..., -0.0037, -0.0036, -0.0037], + [-0.0454, -0.0437, -0.0432, ..., 0.0051, 0.0052, 0.0051], + ..., + [-0.0498, -0.0488, -0.0476, ..., 0.0035, 0.0036, 0.0037], + [-0.0439, -0.0432, -0.0425, ..., 0.0037, 0.0040, 0.0040], + [-0.0310, -0.0295, -0.0284, ..., 0.0011, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0058, 0.0006, 0.0046, ..., 0.0025, 0.0046, -0.0006], + [ 0.0018, -0.0023, -0.0004, ..., 0.0001, 0.0035, 0.0018], + [-0.0065, -0.0017, -0.0055, ..., -0.0077, -0.0070, 0.0017], + ..., + [-0.0293, 0.0015, -0.0344, ..., -0.0378, -0.0361, -0.0017], + [-0.0250, -0.0064, -0.0226, ..., -0.0249, -0.0222, 0.0072], + [ 0.0065, -0.0074, 0.0048, ..., 0.0065, -0.0005, 0.0062]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_cross_attn.1.weight': tensor([[ 6.6406e-02, 6.4941e-02, 6.3965e-02, ..., 4.8161e-05, + 1.6093e-05, 5.9843e-05], + [ 4.4556e-03, 4.4250e-03, 4.4250e-03, ..., -3.4332e-05, + -1.1063e-04, -2.2602e-04], + [-2.2827e-02, -2.1973e-02, -2.0996e-02, ..., -6.8283e-04, + -6.8283e-04, -6.9427e-04], + ..., + [-2.7344e-02, -2.6978e-02, -2.6245e-02, ..., -4.1580e-04, + -3.7575e-04, -3.3760e-04], + [-2.5391e-02, -2.4902e-02, -2.4170e-02, ..., -2.0599e-04, + -1.6022e-04, -2.1267e-04], + [ 5.8838e-02, 5.7617e-02, 5.6885e-02, ..., -1.4591e-04, + -2.6822e-05, -1.7166e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0094, -0.0076, -0.0112, ..., -0.0010, 0.0064, -0.0079], + [ 0.0042, 0.0039, 0.0047, ..., -0.0049, -0.0038, 0.0056], + [ 0.0130, -0.0038, -0.0038, ..., 0.0047, -0.0058, 0.0097], + ..., + [ 0.0679, 0.0107, -0.0051, ..., 0.0074, -0.0149, 0.0483], + [ 0.0288, -0.0056, -0.0077, ..., -0.0071, -0.0209, 0.0310], + [-0.0078, 0.0099, -0.0190, ..., 0.0144, 0.0126, -0.0093]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_mlp.1.weight': tensor([[-0.0043, -0.0045, -0.0044, ..., 0.0003, 0.0003, 0.0003], + [ 0.0659, 0.0640, 0.0635, ..., -0.0038, -0.0039, -0.0038], + [ 0.0415, 0.0408, 0.0403, ..., -0.0039, -0.0039, -0.0039], + ..., + [ 0.0320, 0.0309, 0.0306, ..., -0.0021, -0.0022, -0.0021], + [-0.0124, -0.0114, -0.0105, ..., 0.0014, 0.0014, 0.0013], + [ 0.0254, 0.0253, 0.0253, ..., -0.0025, -0.0026, -0.0026]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_mlp.2.weight': tensor([[-0.0028, 0.0018, 0.0028, ..., 0.0026, 0.0021, 0.0024], + [-0.0002, 0.0020, 0.0021, ..., 0.0019, -0.0001, 0.0010], + [ 0.0019, 0.0001, -0.0037, ..., -0.0031, -0.0019, -0.0029], + ..., + [-0.0029, 0.0013, -0.0189, ..., -0.0081, 0.0028, -0.0057], + [ 0.0151, -0.0075, -0.0244, ..., -0.0165, -0.0144, -0.0148], + [ 0.0172, -0.0195, -0.0133, ..., -0.0166, -0.0178, -0.0162]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.after_proj.weight': tensor([[ 0.0104, -0.0075, 0.0262, ..., -0.0078, 0.0193, -0.0015], + [-0.0023, -0.0077, 0.0088, ..., 0.0153, 0.0086, -0.0063], + [ 0.0081, 0.0028, -0.0168, ..., 0.0070, 0.0148, -0.0087], + ..., + [-0.0050, 0.0074, 0.0249, ..., -0.0206, -0.0128, -0.0078], + [-0.0036, 0.0177, 0.0115, ..., -0.0117, 0.0003, 0.0161], + [-0.0123, -0.0178, -0.0145, ..., 0.0075, 0.0110, -0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.after_proj.bias': tensor([-4.2419e-03, 5.1880e-03, 4.8523e-03, ..., 6.7749e-03, + 7.0572e-05, -1.3885e-03], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_proj.weight': tensor([[ 0.0249, -0.0265, -0.0237, ..., 0.0074, -0.0041, 0.0330], + [ 0.0048, 0.0044, -0.0016, ..., -0.0142, 0.0068, -0.0049], + [ 0.0253, 0.0031, -0.0038, ..., 0.0045, -0.0076, 0.0183], + ..., + [-0.0087, -0.0070, 0.0021, ..., -0.0044, -0.0036, -0.0004], + [ 0.0315, 0.0259, -0.0068, ..., -0.0079, -0.0048, 0.0042], + [-0.0124, -0.0121, 0.0179, ..., -0.0028, 0.0063, -0.0244]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_norm.weight': tensor([ 1.3750e+00, 5.2344e-01, 6.6406e-01, 4.7852e-01, 1.3359e+00, + 1.4893e-02, 9.6436e-03, 1.9789e-05, 2.1057e-03, 1.5137e-02, + 3.3677e-06, -7.6252e-09, 1.1406e+00, -2.5988e-05, -3.5508e-18, + -4.7119e-02, 1.1597e-02, 2.6489e-02, 2.6703e-04, 1.0010e-02, + 1.5527e-01, -3.7909e-05, 9.6875e-01, 1.5938e+00, 7.1094e-01, + 1.1016e+00, 7.1094e-01, 1.9141e-01, 2.3682e-02, -1.4099e-02, + 7.1106e-03, 3.1090e-04, 7.5989e-03, -3.0212e-03, -1.0071e-03, + 2.9663e-02, 1.3965e-01, -2.5272e-05, 5.9814e-03, 2.1973e-03, + -7.0930e-06, -6.2883e-06, -1.7956e-06, 1.5625e+00, 6.4844e-01, + 1.5000e+00, 1.0625e+00, 4.8633e-01, -3.2227e-02, -3.9673e-03, + -9.9182e-04, 2.7222e-02, 1.3184e-01, -3.9101e-05, 6.4373e-05, + -4.3213e-06, -2.4033e-04, -5.0659e-03, -5.5176e-02, -3.1832e-10, + 3.2471e-02, -1.6809e-05, -2.2461e-02, -2.8320e-02, 8.9844e-01, + 1.2422e+00, 1.3438e+00, 1.3828e+00, 6.1328e-01, -2.0508e-02, + 9.5215e-03, 8.8215e-05, 1.2024e-02, 1.2283e-03, 9.0525e-07, + 1.7695e-07, 6.6895e-02, 9.3460e-05, 9.1903e-20, 1.0000e+00, + 1.1797e+00, 2.8564e-02, 2.2507e-04, 3.1494e-02, 1.1475e-01, + 1.1475e-02, 1.6719e+00, 9.8438e-01, 1.4609e+00, 1.5391e+00, + 2.9102e-01, 6.1328e-01, 5.2490e-02, -8.4229e-03, -1.1719e-02, + 1.7357e-04, -3.0396e-02, -1.8677e-02, -3.6316e-03, 4.1260e-02, + 1.3281e-01, -4.4584e-05, 1.1875e+00, -3.6478e-05, -9.1195e-06, + -9.1553e-05, -5.9372e-08, 1.1875e+00, 1.6328e+00, 1.5469e+00, + 1.5703e+00, 8.1250e-01, 1.9775e-02, -2.6001e-02, 3.2959e-03, + 3.4424e-02, 1.3770e-01, 1.4544e-05, 4.1723e-05, -3.1441e-06, + -4.1723e-06, -6.1512e-05, 1.2266e+00, -8.3674e-11, 3.0640e-02, + -5.6028e-06, 9.6484e-01, 1.2188e+00], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.self_attn.k_proj.weight': tensor([[-0.0034, 0.0308, 0.0071, ..., 0.0121, 0.0184, 0.0039], + [ 0.0065, 0.0037, 0.0056, ..., -0.0039, 0.0052, 0.0047], + [-0.0232, 0.0197, 0.0015, ..., 0.0042, 0.0145, 0.0150], + ..., + [ 0.0044, -0.0096, -0.0056, ..., 0.0079, 0.0087, 0.0111], + [ 0.0140, 0.0077, 0.0138, ..., -0.0079, -0.0084, 0.0176], + [ 0.0022, -0.0189, 0.0095, ..., -0.0082, 0.0088, 0.0299]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.k_norm.weight': tensor([ 1.4141e+00, 7.5781e-01, 4.7070e-01, 4.7461e-01, 1.3438e+00, + 1.5991e-02, 1.3977e-02, 8.8692e-05, 7.1716e-03, 1.5076e-02, + 3.2634e-06, -4.3656e-09, 1.1406e+00, 4.3869e-05, -3.5779e-18, + -3.7537e-03, -1.1206e-04, 9.2773e-02, 3.2196e-03, -3.2196e-03, + 4.0283e-03, -3.9339e-05, 1.0625e+00, 1.4609e+00, 4.9609e-01, + 1.2266e+00, 7.8125e-01, 2.9102e-01, 1.2451e-02, 5.8289e-03, + 1.0254e-02, 3.3379e-04, 8.3008e-03, 4.3213e-02, 1.8539e-03, + 9.0332e-03, 2.7710e-02, 1.7643e-05, 2.1729e-02, 2.1973e-03, + -8.8811e-06, -2.3395e-06, 1.6317e-06, 1.7188e+00, 1.1875e+00, + 1.5156e+00, 1.2734e+00, 4.8828e-01, -1.6357e-02, 6.4453e-02, + 3.1281e-03, -9.9487e-03, 4.7913e-03, -3.2902e-05, -1.7047e-05, + 3.9935e-06, -2.5177e-04, -5.0659e-03, 3.0640e-02, 2.7831e-10, + -3.9673e-03, -1.6212e-05, -3.0029e-02, -2.9907e-02, 6.8359e-01, + 1.1406e+00, 1.3750e+00, 1.3594e+00, 4.8242e-01, -1.9165e-02, + 6.1035e-03, -4.5300e-06, 3.5553e-03, -9.6893e-04, 1.1921e-06, + 1.8068e-07, 5.3406e-03, 5.8889e-05, 6.7424e-19, 1.0000e+00, + 1.1797e+00, 8.1055e-02, -4.8828e-04, -3.8300e-03, -3.6316e-03, + 1.1597e-02, 1.5859e+00, 1.2266e+00, 1.5703e+00, 1.5703e+00, + 8.7402e-02, 6.0547e-01, 2.3193e-02, 1.3977e-02, -7.9346e-03, + 1.0395e-04, 3.1738e-02, 6.0303e-02, 7.0801e-03, -1.0254e-02, + -2.5879e-02, -5.1022e-05, 1.1875e+00, 3.6478e-05, -2.4199e-05, + -9.1076e-05, -6.9384e-08, 1.0469e+00, 1.4766e+00, 1.4609e+00, + 1.6797e+00, 8.2031e-01, -2.5024e-02, 9.8633e-02, 3.6163e-03, + 1.9165e-02, -1.6602e-02, -2.7657e-05, 3.2425e-05, -2.9951e-06, + -2.0266e-06, 7.4387e-05, 1.2266e+00, -2.8081e-11, -1.0742e-02, + 5.3346e-06, 9.6484e-01, 1.2188e+00], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.self_attn.v_proj.weight': tensor([[-0.0051, 0.0036, -0.0222, ..., -0.0315, 0.0049, 0.0051], + [ 0.0030, 0.0017, -0.0138, ..., 0.0074, 0.0099, 0.0056], + [-0.0173, -0.0101, 0.0111, ..., 0.0151, -0.0081, -0.0086], + ..., + [ 0.0072, 0.0109, -0.0194, ..., -0.0356, 0.0075, 0.0027], + [ 0.0049, 0.0150, -0.0041, ..., 0.0060, 0.0060, -0.0049], + [ 0.0119, -0.0081, -0.0173, ..., 0.0171, -0.0041, 0.0049]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.output_proj.weight': tensor([[ 0.0085, 0.0261, -0.0267, ..., 0.0055, -0.0234, 0.0265], + [-0.0041, -0.0102, -0.0131, ..., 0.0220, 0.0256, -0.0087], + [ 0.0023, 0.0040, -0.0027, ..., -0.0017, 0.0041, 0.0084], + ..., + [ 0.0037, 0.0060, -0.0251, ..., 0.0096, 0.0140, -0.0101], + [-0.0144, 0.0079, 0.0062, ..., 0.0143, 0.0064, -0.0052], + [ 0.0088, -0.0034, -0.0133, ..., 0.0095, 0.0078, 0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_proj.weight': tensor([[-0.0112, 0.0037, -0.0019, ..., -0.0033, -0.0048, -0.0020], + [ 0.0121, 0.0047, 0.0027, ..., -0.0041, -0.0039, 0.0050], + [-0.0042, -0.0073, -0.0038, ..., -0.0074, 0.0123, 0.0027], + ..., + [-0.0063, -0.0048, -0.0037, ..., 0.0152, -0.0041, -0.0047], + [-0.0011, -0.0074, 0.0025, ..., 0.0134, -0.0055, -0.0047], + [ 0.0025, 0.0034, 0.0023, ..., -0.0037, 0.0043, 0.0037]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_norm.weight': tensor([-6.5613e-04, 1.3184e-01, 1.4526e-02, 6.0547e-01, 2.2339e-02, + 1.2634e-02, 8.5449e-03, 3.2806e-03, 7.0190e-03, 1.1279e-01, + 1.8677e-02, 5.0049e-02, 2.6367e-01, 1.3977e-02, 1.7090e-01, + 1.3245e-02, 5.3467e-02, 3.1494e-02, 1.6937e-03, 1.7578e-02, + 8.6060e-03, 1.8262e-01, 2.7344e-01, 1.8555e-01, 2.0752e-02, + 1.6016e-01, 5.5908e-02, 9.7656e-03, 4.0234e-01, 6.2866e-03, + 1.1292e-02, 5.0781e-01, 5.5664e-02, 3.3936e-02, 1.6113e-02, + 1.3123e-02, 6.7871e-02, 4.5508e-01, 5.3906e-01, 1.4954e-02, + 3.8910e-03, 1.0449e-01, 5.2490e-03, 5.3906e-01, 1.8188e-02, + 4.3213e-02, 1.0437e-02, 3.9258e-01, 7.2327e-03, 1.7944e-02, + 6.8359e-02, 1.8799e-02, 1.0437e-02, 5.0391e-01, -8.5831e-05, + 1.9897e-02, 5.5664e-02, 7.7148e-02, 9.0942e-03, 1.0132e-02, + 5.7983e-03, 9.6436e-03, 1.4099e-02, 3.3008e-01, 2.0264e-02, + 2.8534e-03, 2.6001e-02, 5.2979e-02, 6.2988e-02, 1.6235e-02, + 6.1951e-03, 6.8359e-01, -3.4094e-05, 6.1328e-01, 9.0332e-03, + 1.5198e-02, 2.2217e-02, 2.6953e-01, 3.2471e-02, 9.1171e-04, + 1.0645e-01, 9.7656e-03, 4.0771e-02, 1.1169e-02, 2.1362e-02, + 7.7209e-03, 2.9602e-03, 6.9336e-02, 1.9434e-01, 8.9111e-03, + 9.3994e-03, 1.5625e-02, 4.6875e-02, 2.8320e-02, 2.3193e-02, + 2.3560e-02, 1.4709e-02, 1.6846e-02, 6.2012e-02, 1.9287e-02, + 3.1250e-02, 7.2632e-03, 1.0803e-02, 5.4932e-02, 6.5234e-01, + 6.8848e-02, 2.6550e-03, 4.4531e-01, 1.0925e-02, 1.8799e-02, + 1.6479e-02, 1.5015e-02, 8.4043e-06, 8.9844e-02, 4.2969e-01, + 2.4414e-02, 1.7334e-02, 6.2109e-01, 1.1914e-01, 4.0039e-02, + 1.2146e-02, 7.1875e-01, 1.2891e-01, 1.8799e-02, 1.0925e-02, + 4.4250e-03, 2.0752e-02, 1.3046e-03], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_proj.weight': tensor([[-0.0046, -0.0034, 0.0103, ..., 0.0059, -0.0056, 0.0032], + [-0.0042, 0.0003, -0.0034, ..., 0.0005, -0.0081, 0.0029], + [-0.0002, -0.0016, -0.0085, ..., 0.0123, -0.0168, -0.0072], + ..., + [ 0.0040, 0.0046, -0.0043, ..., 0.0052, 0.0020, -0.0061], + [ 0.0093, 0.0029, -0.0031, ..., -0.0035, 0.0093, 0.0023], + [-0.0056, 0.0007, -0.0092, ..., -0.0081, -0.0060, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_norm.weight': tensor([ 6.8665e-04, 1.3184e-01, 1.4038e-02, 6.0156e-01, 2.1851e-02, + 1.2573e-02, 8.7280e-03, 3.2654e-03, 7.4158e-03, 1.1279e-01, + 1.8677e-02, 4.8584e-02, 2.6367e-01, 1.3855e-02, 1.7090e-01, + 1.3916e-02, 5.2734e-02, 3.1738e-02, 1.6327e-03, 1.7456e-02, + 8.9111e-03, 1.8359e-01, 2.7539e-01, 1.8555e-01, 2.0508e-02, + 1.6211e-01, 5.4443e-02, 7.9956e-03, 4.0430e-01, 6.2561e-03, + 1.1536e-02, 5.0781e-01, 5.5664e-02, 3.3936e-02, 1.6479e-02, + 1.2451e-02, 6.8359e-02, 4.5508e-01, 5.3906e-01, 1.4954e-02, + 3.8910e-03, 1.0449e-01, 5.2185e-03, 5.3906e-01, 1.8433e-02, + 4.3213e-02, 1.0803e-02, 3.9258e-01, 7.3853e-03, 1.7822e-02, + 6.8359e-02, 1.9043e-02, 1.0376e-02, 5.0391e-01, 1.2589e-04, + 1.9897e-02, 5.5664e-02, 7.7148e-02, 9.0942e-03, 8.3008e-03, + 5.7373e-03, 9.6436e-03, 1.4404e-02, 3.3008e-01, 1.9775e-02, + 2.9755e-03, 2.5513e-02, 5.2979e-02, 6.2988e-02, 1.6235e-02, + 6.5918e-03, 6.7969e-01, -3.3855e-05, 6.1328e-01, 9.2163e-03, + -1.4771e-02, 2.2217e-02, 2.6953e-01, 3.3203e-02, 9.1934e-04, + 1.0693e-01, 9.6436e-03, 4.0039e-02, 1.0986e-02, 2.0874e-02, + 7.7820e-03, 3.9062e-03, 6.9824e-02, 1.9434e-01, 8.9111e-03, + 9.5825e-03, 1.4709e-02, 4.7119e-02, 2.8442e-02, 2.4048e-02, + 2.3682e-02, 1.4709e-02, 1.6968e-02, 6.2012e-02, 1.9165e-02, + 3.1250e-02, 7.4768e-03, 1.1230e-02, 5.5420e-02, 6.4844e-01, + 6.8848e-02, 2.6855e-03, 4.4727e-01, 1.0620e-02, 1.8677e-02, + 1.6968e-02, 1.5015e-02, 1.1623e-05, 8.9355e-02, 4.2969e-01, + 2.4170e-02, 1.7212e-02, 6.2109e-01, 1.1914e-01, 4.0283e-02, + 1.2085e-02, 7.2266e-01, 1.2793e-01, 1.7822e-02, 1.0681e-02, + 4.3640e-03, 2.0630e-02, 7.3624e-04], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.v_proj.weight': tensor([[ 0.0170, -0.0073, -0.0155, ..., 0.0033, -0.0150, -0.0077], + [-0.0186, -0.0008, -0.0024, ..., -0.0024, -0.0233, 0.0194], + [-0.0188, -0.0067, -0.0031, ..., -0.0038, 0.0283, 0.0312], + ..., + [-0.0130, -0.0038, -0.0086, ..., -0.0001, 0.0154, 0.0029], + [ 0.0229, 0.0045, 0.0222, ..., 0.0052, 0.0056, 0.0077], + [ 0.0139, 0.0012, 0.0033, ..., 0.0089, -0.0164, -0.0210]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0003, 0.0027, ..., -0.0012, 0.0021, 0.0156], + [ 0.0109, -0.0014, 0.0027, ..., 0.0042, 0.0002, -0.0014], + [ 0.0042, 0.0008, -0.0021, ..., -0.0047, -0.0039, -0.0054], + ..., + [-0.0178, -0.0018, 0.0037, ..., 0.0044, -0.0113, 0.0034], + [ 0.0046, -0.0100, 0.0056, ..., -0.0310, 0.0078, -0.0223], + [ 0.0013, -0.0036, 0.0085, ..., 0.0297, -0.0044, 0.0217]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0064, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [-0.0015, 0.0150, -0.0159, ..., -0.0041, -0.0055, 0.0679], + ..., + [-0.0208, 0.0130, -0.0273, ..., 0.0308, -0.0072, -0.0068], + [-0.0337, -0.0025, 0.0054, ..., -0.0040, 0.0172, 0.0231], + [-0.0255, 0.0040, 0.0030, ..., 0.0408, 0.0430, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0031], + [ 0.0198, -0.0208, 0.0130, ..., 0.0089, -0.0337, -0.0055], + [ 0.0302, 0.0101, -0.0023, ..., -0.0304, 0.0199, 0.0024], + ..., + [ 0.0171, -0.0275, -0.0075, ..., 0.0231, 0.0182, -0.0039], + [ 0.0221, 0.0388, 0.0056, ..., 0.0042, -0.0488, -0.0361], + [-0.0005, 0.0120, -0.0051, ..., 0.0037, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img.weight': tensor([[ 0.0273, -0.0041, 0.0118, ..., 0.0012, 0.0347, 0.0615], + [-0.0258, 0.0042, -0.0197, ..., 0.0139, -0.0047, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0097, 0.0132, 0.0184, ..., 0.0110, 0.0087, 0.0515], + [-0.0006, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.mlp.layer1.weight': tensor([[ 0.0017, -0.0165, 0.0046, ..., 0.0017, -0.0049, -0.0064], + [-0.0162, 0.0078, -0.0029, ..., 0.0124, 0.0065, -0.0203], + [-0.0122, -0.0107, 0.0084, ..., -0.0070, -0.0010, -0.0610], + ..., + [-0.0042, 0.0038, 0.0038, ..., 0.0156, 0.0085, -0.0038], + [-0.0093, 0.0140, -0.0031, ..., -0.0092, 0.0054, -0.0396], + [-0.0037, 0.0162, -0.0016, ..., 0.0114, 0.0172, -0.0091]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.mlp.layer2.weight': tensor([[ 0.0040, -0.0082, -0.0236, ..., 0.0062, -0.0112, -0.0124], + [-0.0309, -0.0120, -0.0055, ..., -0.0011, -0.0064, -0.0074], + [ 0.0049, -0.0071, -0.0030, ..., 0.0031, -0.0033, 0.0087], + ..., + [-0.0028, -0.0079, -0.0432, ..., -0.0063, -0.0098, -0.0110], + [-0.0104, 0.0050, 0.0120, ..., 0.0024, 0.0128, -0.0052], + [-0.0078, -0.0063, -0.0037, ..., -0.0096, 0.0097, -0.0388]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0396, 0.0398, 0.0381, ..., -0.0042, -0.0042, -0.0044], + [-0.0383, -0.0369, -0.0359, ..., 0.0043, 0.0048, 0.0047], + [-0.0400, -0.0393, -0.0383, ..., 0.0045, 0.0049, 0.0049], + ..., + [-0.0422, -0.0408, -0.0400, ..., 0.0022, 0.0022, 0.0024], + [-0.0415, -0.0408, -0.0398, ..., 0.0021, 0.0022, 0.0022], + [-0.0039, -0.0039, -0.0039, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_self_attn.2.weight': tensor([[ 3.8757e-03, -3.4637e-03, -3.5248e-03, ..., 2.0504e-05, + -5.6744e-05, 5.4016e-03], + [ 3.3722e-03, -4.0283e-03, -3.2654e-03, ..., -4.7302e-03, + -4.6997e-03, -1.7776e-03], + [ 4.8218e-03, -6.3171e-03, -5.7678e-03, ..., -3.7842e-03, + -3.8147e-03, 1.3123e-03], + ..., + [ 3.9062e-02, -4.1992e-02, -4.2480e-02, ..., -2.0630e-02, + -1.7334e-02, 1.6235e-02], + [ 3.9978e-03, -5.7983e-03, -5.8899e-03, ..., 3.8910e-03, + 3.8910e-03, 5.5847e-03], + [-1.4038e-02, 1.5564e-02, 1.4099e-02, ..., 1.7090e-02, + 1.6968e-02, 1.4526e-02]], dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9570e-02, 5.8838e-02, 5.7617e-02, ..., -1.1015e-04, + -3.9339e-05, -1.9169e-04], + [ 6.3965e-02, 6.2012e-02, 5.9326e-02, ..., 3.4904e-04, + 4.6158e-04, 4.9210e-04], + [-4.8340e-02, -4.6387e-02, -4.5654e-02, ..., -2.6464e-05, + 2.8014e-05, -8.0109e-05], + ..., + [ 3.1494e-02, 3.0518e-02, 2.9785e-02, ..., 1.1206e-04, + 4.5061e-05, 7.3433e-05], + [ 6.1035e-02, 5.9326e-02, 5.8594e-02, ..., 5.2643e-04, + 5.6076e-04, 5.5313e-04], + [-4.7852e-02, -4.6143e-02, -4.5166e-02, ..., -3.2234e-04, + -3.6430e-04, -3.2806e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0149, 0.0073, -0.0080, ..., 0.0018, -0.0047, -0.0041], + [-0.0178, 0.0121, 0.0144, ..., -0.0096, -0.0124, 0.0083], + [ 0.0011, 0.0030, -0.0011, ..., 0.0009, 0.0003, -0.0002], + ..., + [ 0.0260, 0.0184, -0.0149, ..., 0.0078, 0.0080, 0.0037], + [ 0.0075, 0.0334, -0.0057, ..., 0.0076, 0.0325, -0.0178], + [-0.0044, 0.0203, 0.0051, ..., -0.0114, -0.0110, 0.0240]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_mlp.1.weight': tensor([[-0.0986, -0.0967, -0.0942, ..., 0.0031, 0.0034, 0.0033], + [-0.0371, -0.0352, -0.0339, ..., 0.0008, 0.0008, 0.0007], + [ 0.0566, 0.0549, 0.0544, ..., -0.0017, -0.0017, -0.0017], + ..., + [ 0.0364, 0.0354, 0.0347, ..., -0.0018, -0.0018, -0.0016], + [ 0.0466, 0.0444, 0.0430, ..., -0.0010, -0.0009, -0.0010], + [-0.0518, -0.0505, -0.0496, ..., 0.0022, 0.0022, 0.0021]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_mlp.2.weight': tensor([[-0.0228, -0.0165, 0.0110, ..., -0.0018, 0.0181, -0.0115], + [ 0.0017, -0.0033, -0.0016, ..., -0.0016, 0.0030, 0.0016], + [-0.0005, -0.0007, -0.0008, ..., -0.0010, 0.0007, 0.0009], + ..., + [ 0.0454, -0.0031, -0.0184, ..., -0.0126, 0.0032, 0.0193], + [ 0.0459, -0.0288, -0.0371, ..., -0.0337, 0.0286, 0.0366], + [ 0.0128, -0.0366, -0.0251, ..., -0.0284, 0.0374, 0.0240]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.after_proj.weight': tensor([[-0.0054, -0.0084, -0.0073, ..., 0.0059, 0.0018, -0.0046], + [ 0.0066, 0.0057, 0.0067, ..., 0.0050, 0.0052, 0.0131], + [-0.0096, -0.0041, -0.0051, ..., -0.0023, -0.0037, 0.0110], + ..., + [-0.0015, -0.0067, -0.0140, ..., -0.0055, -0.0112, 0.0187], + [-0.0049, -0.0150, -0.0076, ..., 0.0105, -0.0118, -0.0096], + [-0.0036, 0.0160, 0.0104, ..., -0.0038, 0.0221, -0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.after_proj.bias': tensor([-3.4790e-03, 2.1820e-03, 1.1139e-03, ..., 4.6387e-03, + 9.3937e-05, -4.8637e-04], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_proj.weight': tensor([[-0.0110, 0.0153, 0.0214, ..., 0.0090, 0.0057, -0.0190], + [ 0.0203, -0.0014, 0.0074, ..., 0.0058, -0.0027, 0.0075], + [-0.0127, -0.0063, 0.0129, ..., -0.0188, -0.0093, -0.0134], + ..., + [ 0.0065, 0.0038, 0.0002, ..., 0.0057, 0.0030, 0.0051], + [ 0.0098, -0.0222, -0.0029, ..., -0.0398, -0.0059, -0.0120], + [-0.0065, -0.0048, -0.0052, ..., 0.0099, 0.0019, 0.0074]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_norm.weight': tensor([ 1.3672e+00, 1.8047e+00, 1.3828e+00, 1.5859e+00, 2.2852e-01, + 5.6396e-02, 6.5308e-03, 4.2915e-06, -1.7694e-33, 3.5477e-04, + 3.9062e-02, 1.2695e-01, 1.0620e-02, 3.2959e-02, -1.1719e-02, + 4.9744e-03, 6.3965e-02, 9.5703e-01, 8.8281e-01, 1.0376e-03, + 8.9169e-05, 1.2734e+00, 2.2188e+00, 1.9609e+00, 1.0469e+00, + 1.1094e+00, 1.5312e+00, 8.9264e-04, -1.6689e-04, -5.9128e-04, + 1.1292e-02, 1.7944e-02, 9.9487e-03, 1.0156e+00, 9.1553e-03, + 1.0156e+00, 7.3242e-04, -6.8665e-05, 7.4768e-03, -1.2024e-02, + 1.9670e-06, -5.9605e-06, 6.8665e-05, 2.3125e+00, 1.5000e+00, + 1.5781e+00, 1.6016e+00, 1.7578e+00, 1.1523e-01, 7.4768e-04, + 2.4048e-02, 2.4707e-01, 1.1172e+00, 1.3281e-01, 1.4038e-03, + 2.1973e-01, 3.0327e-04, 1.2305e-01, 6.1035e-05, -4.4861e-03, + -7.4463e-03, -5.6152e-03, -6.3477e-02, 1.1328e+00, 1.9609e+00, + 8.3984e-01, 1.1094e+00, 1.3281e+00, 7.7637e-02, -2.1484e-02, + 1.2573e-02, 2.5511e-05, -2.9491e-33, -1.1597e-03, 5.2795e-03, + 1.0559e-02, 1.4496e-03, -3.5156e-02, -2.8442e-02, 1.7944e-02, + 8.8867e-02, 3.6133e-02, 1.0254e-02, 6.1951e-03, -3.7994e-03, + 2.8687e-03, 1.0156e+00, 1.1953e+00, 1.6094e+00, 1.5703e+00, + 9.3750e-01, -7.4768e-03, 5.8365e-04, -2.9297e-03, 3.1281e-03, + -6.8665e-03, 2.1729e-02, 5.6152e-02, 1.7090e-02, 4.4678e-02, + -2.8229e-03, -1.3199e-03, 1.5234e-01, 4.2439e-05, 5.4550e-04, + -8.8811e-06, 2.8534e-03, 8.0469e-01, 1.7734e+00, 1.5547e+00, + 1.1016e+00, 1.2031e+00, 6.2988e-02, 6.5994e-04, 4.1992e-02, + 1.2188e+00, 1.1035e-01, 1.1641e+00, -2.0790e-04, 1.3794e-02, + 1.1475e-02, 5.6641e-02, 2.0313e-04, -2.2949e-02, -5.6028e-05, + 1.2695e-02, 1.0234e+00, -9.7656e-03], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.self_attn.k_proj.weight': tensor([[-0.0220, -0.0096, -0.0002, ..., -0.0080, 0.0030, -0.0188], + [ 0.0066, -0.0347, -0.0034, ..., 0.0116, 0.0139, 0.0144], + [-0.0121, 0.0195, -0.0061, ..., -0.0107, -0.0099, 0.0030], + ..., + [ 0.0010, 0.0106, -0.0084, ..., -0.0008, 0.0042, 0.0038], + [ 0.0222, -0.0081, 0.0102, ..., -0.0147, -0.0076, 0.0221], + [-0.0123, -0.0031, -0.0052, ..., 0.0056, 0.0062, 0.0053]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.k_norm.weight': tensor([ 9.6094e-01, 1.7266e+00, 1.3984e+00, 1.4844e+00, 2.3535e-01, + 6.2500e-02, -1.8845e-03, 1.1861e-05, -1.5167e-33, 2.5177e-04, + 3.9551e-02, 1.2793e-01, -1.1780e-02, 5.1025e-02, -1.1902e-02, + -4.7607e-03, 1.0925e-02, 9.5703e-01, 8.8281e-01, -9.4986e-04, + 2.2316e-04, 1.2734e+00, 1.8828e+00, 1.7734e+00, 4.3945e-02, + 5.3906e-01, 1.5469e+00, -6.9580e-03, -3.4714e-04, 4.4823e-04, + -1.1597e-02, 1.7822e-02, -6.1340e-03, 1.0078e+00, -7.6599e-03, + 1.0078e+00, 7.4768e-04, -1.0681e-04, -9.5215e-03, -1.2146e-02, + 2.9206e-06, -7.8082e-06, 6.4373e-05, 1.8672e+00, 1.0234e+00, + 1.5391e+00, 1.6641e+00, 1.7109e+00, 7.5684e-02, 8.0872e-04, + -3.9368e-03, -6.7444e-03, 1.1172e+00, 1.7188e-01, 1.3351e-03, + 2.1973e-01, -8.8501e-04, 1.2305e-01, -1.3173e-05, 7.1289e-02, + -7.4768e-03, 2.4414e-02, -7.1777e-02, 1.1328e+00, 2.0781e+00, + 1.2109e+00, 9.2969e-01, 1.3828e+00, -1.7944e-02, -1.0254e-02, + -1.6968e-02, -2.2888e-05, 4.3815e-33, 1.3275e-03, 1.7456e-02, + 4.6143e-02, 6.2561e-04, 4.9561e-02, -3.9795e-02, -1.7944e-02, + -5.6152e-02, 2.8809e-02, 1.3062e-02, 6.1340e-03, 3.9673e-03, + -1.1215e-03, 1.5234e+00, 1.5625e+00, 1.6953e+00, 1.6172e+00, + 5.4297e-01, 5.5542e-03, -4.3297e-04, -2.9907e-03, 9.4223e-04, + 7.1411e-03, -3.7384e-03, 5.9509e-03, -1.6235e-02, 8.3008e-03, + -2.8076e-03, -1.3275e-03, 1.5234e-01, 3.1233e-05, 5.6839e-04, + -1.0788e-05, 2.8534e-03, 1.6250e+00, 1.9844e+00, 1.5469e+00, + 5.5859e-01, 1.2969e+00, -4.3030e-03, 2.6512e-04, -3.1738e-02, + 1.2188e+00, 1.5442e-02, 1.1641e+00, -2.6512e-04, 1.6724e-02, + 1.1719e-02, 5.5908e-02, 2.0313e-04, 5.7617e-02, -1.0014e-04, + 2.2949e-02, 1.0234e+00, 3.2959e-03], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.self_attn.v_proj.weight': tensor([[ 0.0020, -0.0152, 0.0064, ..., 0.0461, -0.0211, 0.0038], + [ 0.0127, 0.0267, -0.0182, ..., 0.0046, 0.0138, -0.0134], + [-0.0003, 0.0297, -0.0025, ..., -0.0084, 0.0220, 0.0025], + ..., + [ 0.0086, 0.0102, -0.0201, ..., 0.0080, 0.0078, 0.0047], + [-0.0018, 0.0332, 0.0127, ..., 0.0145, 0.0266, -0.0056], + [-0.0086, -0.0042, 0.0227, ..., 0.0187, 0.0172, -0.0100]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.output_proj.weight': tensor([[ 0.0066, 0.0055, 0.0227, ..., 0.0034, 0.0201, -0.0080], + [ 0.0229, -0.0079, 0.0054, ..., -0.0186, 0.0041, -0.0114], + [ 0.0059, 0.0130, -0.0233, ..., -0.0077, -0.0072, 0.0134], + ..., + [-0.0129, -0.0464, -0.0078, ..., -0.0089, -0.0006, -0.0193], + [ 0.0032, 0.0084, -0.0201, ..., 0.0087, 0.0236, 0.0068], + [-0.0566, -0.0126, 0.0322, ..., 0.0216, -0.0221, -0.0166]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_proj.weight': tensor([[ 0.0098, -0.0079, 0.0138, ..., 0.0038, 0.0074, 0.0201], + [ 0.0078, -0.0058, -0.0052, ..., -0.0090, -0.0052, 0.0119], + [-0.0033, 0.0165, -0.0048, ..., 0.0044, -0.0076, -0.0115], + ..., + [ 0.0093, 0.0032, -0.0132, ..., 0.0160, 0.0046, -0.0068], + [-0.0110, -0.0153, -0.0032, ..., -0.0090, 0.0082, 0.0144], + [ 0.0095, -0.0047, -0.0076, ..., 0.0019, 0.0171, -0.0103]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_norm.weight': tensor([0.1064, 0.0159, 0.4277, 0.0698, 0.1680, 0.0361, 0.1484, 0.1357, 0.5938, + 0.0364, 0.0618, 0.0195, 0.0491, 0.0767, 0.0566, 0.1963, 0.4844, 0.0320, + 0.7461, 0.0352, 0.0437, 0.1689, 0.0178, 0.0801, 0.0481, 0.4082, 0.5508, + 0.2559, 0.0771, 0.0535, 0.0403, 0.2158, 0.6680, 0.2441, 0.0518, 0.0181, + 0.0105, 0.1099, 0.0156, 0.0476, 0.0493, 0.1289, 0.0182, 0.0400, 0.2041, + 0.0532, 0.3555, 0.0205, 0.0500, 0.0139, 0.0310, 0.3848, 0.5391, 0.3438, + 0.0903, 0.6484, 0.5977, 0.0400, 0.1235, 0.0322, 0.0854, 0.5586, 0.0102, + 0.5508, 0.1719, 0.3926, 0.0825, 0.0508, 0.0625, 0.4688, 0.0242, 0.5234, + 0.0214, 0.0356, 0.1338, 0.0874, 0.0287, 0.0889, 0.6367, 0.0352, 0.0889, + 0.0825, 0.0435, 0.0391, 0.4219, 0.0640, 0.0625, 0.0986, 0.0352, 0.1426, + 0.0459, 0.1758, 0.0537, 0.0588, 0.0295, 0.1221, 0.0131, 0.3613, 0.5820, + 0.0342, 0.0562, 0.2949, 0.0208, 0.1211, 0.0442, 0.2314, 0.0228, 0.6328, + 0.5859, 0.0327, 0.0120, 0.7695, 0.0192, 0.0137, 0.4941, 0.2070, 0.2754, + 0.7500, 0.0410, 0.0757, 0.0776, 0.5781, 0.6406, 0.0125, 0.0272, 0.2021, + 0.0249, 0.0205], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_proj.weight': tensor([[-8.7280e-03, 1.8158e-03, 7.0496e-03, ..., -5.6458e-04, + 2.1973e-02, -2.3346e-03], + [ 7.2327e-03, -6.7711e-05, -4.7302e-03, ..., -2.6398e-03, + -2.5330e-03, 1.0986e-02], + [-1.3123e-02, 7.6904e-03, -1.5747e-02, ..., 2.2888e-03, + 6.5002e-03, 5.9509e-03], + ..., + [ 2.8839e-03, 3.1128e-03, -2.5330e-03, ..., 2.6855e-03, + -7.0801e-03, 1.2390e-02], + [-1.1780e-02, -1.8539e-03, -1.4465e-02, ..., 5.3406e-03, + -9.0332e-03, 2.7466e-03], + [ 1.0010e-02, -1.4038e-03, 2.1820e-03, ..., -1.3855e-02, + 5.8594e-03, -1.8997e-03]], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_norm.weight': tensor([0.1064, 0.0157, 0.4277, 0.0698, 0.1680, 0.0364, 0.1484, 0.1348, 0.5938, + 0.0369, 0.0623, 0.0195, 0.0491, 0.0767, 0.0569, 0.1982, 0.4844, 0.0320, + 0.7422, 0.0349, 0.0435, 0.1680, 0.0178, 0.0796, 0.0481, 0.4043, 0.5508, + 0.2539, 0.0771, 0.0532, 0.0413, 0.2178, 0.6680, 0.2441, 0.0525, 0.0181, + 0.0102, 0.1099, 0.0159, 0.0466, 0.0488, 0.1299, 0.0183, 0.0403, 0.2041, + 0.0532, 0.3555, 0.0206, 0.0498, 0.0137, 0.0315, 0.3848, 0.5352, 0.3438, + 0.0903, 0.6484, 0.6016, 0.0398, 0.1235, 0.0325, 0.0854, 0.5586, 0.0106, + 0.5508, 0.1709, 0.3926, 0.0825, 0.0505, 0.0625, 0.4707, 0.0240, 0.5234, + 0.0212, 0.0356, 0.1338, 0.0869, 0.0286, 0.0898, 0.6367, 0.0347, 0.0879, + 0.0830, 0.0435, 0.0391, 0.4219, 0.0645, 0.0618, 0.0977, 0.0354, 0.1445, + 0.0469, 0.1748, 0.0537, 0.0596, 0.0298, 0.1216, 0.0131, 0.3613, 0.5820, + 0.0337, 0.0569, 0.2949, 0.0206, 0.1211, 0.0442, 0.2314, 0.0225, 0.6328, + 0.5859, 0.0325, 0.0119, 0.7695, 0.0187, 0.0140, 0.4961, 0.2070, 0.2754, + 0.7461, 0.0403, 0.0732, 0.0771, 0.5781, 0.6406, 0.0123, 0.0272, 0.2021, + 0.0261, 0.0210], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.v_proj.weight': tensor([[ 0.0135, -0.0027, -0.0303, ..., -0.0010, -0.0011, -0.0116], + [-0.0089, 0.0164, -0.0017, ..., 0.0066, -0.0045, -0.0234], + [ 0.0212, 0.0015, -0.0043, ..., 0.0004, -0.0052, -0.0410], + ..., + [-0.0072, 0.0032, 0.0109, ..., -0.0109, -0.0275, 0.0041], + [-0.0396, -0.0017, -0.0281, ..., -0.0083, 0.0090, -0.0190], + [-0.0095, -0.0029, 0.0029, ..., -0.0099, 0.0098, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.output_proj.weight': tensor([[ 0.0021, 0.0048, -0.0103, ..., -0.0038, 0.0046, 0.0111], + [-0.0361, 0.0028, -0.0017, ..., 0.0057, -0.0110, -0.0220], + [ 0.0036, -0.0078, -0.0037, ..., 0.0044, 0.0173, -0.0050], + ..., + [ 0.0079, -0.0135, 0.0210, ..., 0.0182, -0.0076, -0.0128], + [-0.0109, -0.0053, 0.0071, ..., 0.0124, -0.0144, 0.0092], + [-0.0041, -0.0201, 0.0154, ..., -0.0038, 0.0083, 0.0167]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0050, -0.0065, ..., 0.0266, -0.0278, 0.0078], + ..., + [-0.0217, -0.0410, 0.0488, ..., 0.0039, -0.0022, -0.0255], + [-0.0032, 0.0066, 0.0034, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0055, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, -0.0019, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0039, 0.0293, ..., 0.0065, 0.0089, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0140, 0.0126, -0.0093], + ..., + [-0.0044, 0.0320, -0.0233, ..., -0.0067, 0.0038, 0.0187], + [ 0.0136, -0.0352, 0.0267, ..., -0.0305, -0.0039, 0.0005], + [-0.0205, 0.0143, 0.0008, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0014, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0058, -0.0537, -0.0282, ..., -0.0109, -0.0146, -0.0023], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, 0.0029, 0.0015], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0024, ..., -0.0103, 0.0073, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.mlp.layer1.weight': tensor([[ 0.0123, 0.0036, 0.0047, ..., 0.0121, -0.0084, -0.0044], + [-0.0125, 0.0001, 0.0205, ..., 0.0079, -0.0267, -0.0065], + [ 0.0048, 0.0181, -0.0098, ..., 0.0228, -0.0131, 0.0217], + ..., + [-0.0101, 0.0042, 0.0571, ..., 0.0067, 0.0117, -0.0193], + [ 0.0087, -0.0047, -0.0078, ..., -0.0101, -0.0037, -0.0166], + [ 0.0046, 0.0154, -0.0159, ..., -0.0304, -0.0219, -0.0162]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.mlp.layer2.weight': tensor([[ 0.0048, 0.0209, -0.0045, ..., -0.0162, -0.0251, 0.0036], + [ 0.0038, 0.0232, 0.0094, ..., -0.0051, 0.0177, 0.0135], + [ 0.0050, -0.0222, 0.0045, ..., -0.0055, -0.0045, 0.0055], + ..., + [-0.0053, 0.0055, 0.0253, ..., -0.0049, -0.0100, 0.0018], + [ 0.0039, 0.0140, 0.0219, ..., -0.0047, 0.0172, 0.0111], + [-0.0066, 0.0059, -0.0039, ..., -0.0128, 0.0103, 0.0061]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0469, -0.0464, -0.0449, ..., 0.0010, 0.0008, 0.0009], + [-0.0386, -0.0381, -0.0371, ..., 0.0027, 0.0027, 0.0027], + [-0.0400, -0.0391, -0.0376, ..., 0.0022, 0.0023, 0.0023], + ..., + [ 0.0381, 0.0371, 0.0359, ..., -0.0036, -0.0037, -0.0036], + [ 0.0327, 0.0317, 0.0306, ..., -0.0031, -0.0030, -0.0030], + [-0.0386, -0.0376, -0.0371, ..., 0.0022, 0.0024, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0152, 0.0145, 0.0137, ..., 0.0220, 0.0248, 0.0148], + [ 0.0016, -0.0010, 0.0036, ..., 0.0025, 0.0036, 0.0003], + [-0.0003, -0.0022, -0.0009, ..., 0.0038, 0.0041, -0.0025], + ..., + [-0.0250, -0.0747, -0.0610, ..., 0.0364, 0.0369, -0.0679], + [ 0.0084, 0.0085, 0.0068, ..., 0.0199, 0.0176, 0.0087], + [ 0.0142, 0.0089, 0.0103, ..., 0.0192, 0.0216, 0.0101]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_cross_attn.1.weight': tensor([[-7.9102e-02, -7.6172e-02, -7.4707e-02, ..., -1.0252e-04, + -1.1730e-04, -1.0729e-04], + [ 3.0029e-02, 2.9297e-02, 2.8809e-02, ..., 9.4995e-07, + 2.6345e-05, 1.4687e-04], + [ 2.8442e-02, 2.7710e-02, 2.6489e-02, ..., 2.7466e-04, + 2.9755e-04, 2.0504e-04], + ..., + [ 6.4453e-02, 6.2988e-02, 6.2012e-02, ..., -1.8978e-04, + -1.8406e-04, -1.1444e-04], + [ 6.6895e-02, 6.4453e-02, 6.2988e-02, ..., -1.7643e-04, + -1.1778e-04, -1.6403e-04], + [-3.6133e-02, -3.5156e-02, -3.4424e-02, ..., -5.2929e-05, + 9.9659e-05, 1.2112e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0038, -0.0112, -0.0028, ..., 0.0201, 0.0288, 0.0041], + [ 0.0071, -0.0090, 0.0116, ..., 0.0005, 0.0060, -0.0019], + [ 0.0011, -0.0026, 0.0036, ..., -0.0082, -0.0055, -0.0093], + ..., + [-0.0141, 0.0030, 0.0330, ..., 0.0052, 0.0139, 0.0052], + [-0.0225, 0.0114, 0.0369, ..., 0.0081, 0.0171, -0.0190], + [ 0.0266, -0.0216, 0.0052, ..., -0.0014, 0.0077, 0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_mlp.1.weight': tensor([[-8.9844e-02, -8.6426e-02, -8.3984e-02, ..., 1.8787e-04, + 3.0899e-04, 2.5368e-04], + [ 2.2461e-02, 2.0264e-02, 1.9409e-02, ..., 1.8787e-04, + 2.6894e-04, 2.6321e-04], + [-9.2285e-02, -9.0332e-02, -8.9844e-02, ..., -4.0245e-04, + -3.2806e-04, -3.6621e-04], + ..., + [-8.2031e-02, -8.0566e-02, -7.9102e-02, ..., 2.1076e-04, + 3.1471e-04, 2.1172e-04], + [-6.1279e-02, -5.9570e-02, -5.8594e-02, ..., 2.2888e-04, + 9.2506e-05, 1.0252e-04], + [ 7.9102e-02, 7.7637e-02, 7.6660e-02, ..., -1.7738e-04, + -2.0599e-04, -1.6880e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_mlp.2.weight': tensor([[ 0.0023, 0.0038, 0.0038, ..., -0.0055, -0.0086, 0.0069], + [-0.0018, -0.0021, 0.0057, ..., -0.0016, 0.0037, 0.0015], + [-0.0010, 0.0045, -0.0019, ..., -0.0009, -0.0012, 0.0011], + ..., + [-0.0413, 0.0198, -0.0593, ..., -0.0295, -0.0106, 0.0225], + [ 0.0091, -0.0100, 0.0258, ..., 0.0067, 0.0037, -0.0063], + [ 0.0154, 0.0053, 0.0056, ..., 0.0135, 0.0070, -0.0107]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.after_proj.weight': tensor([[-0.0056, 0.0069, 0.0043, ..., 0.0067, 0.0135, -0.0127], + [-0.0050, -0.0069, 0.0068, ..., -0.0057, 0.0049, -0.0186], + [ 0.0104, -0.0056, 0.0124, ..., 0.0151, 0.0187, 0.0063], + ..., + [-0.0077, -0.0034, 0.0146, ..., 0.0031, -0.0150, -0.0096], + [-0.0122, -0.0050, -0.0278, ..., -0.0188, 0.0081, -0.0032], + [-0.0042, 0.0045, 0.0044, ..., -0.0145, 0.0129, 0.0003]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.after_proj.bias': tensor([ 0.0081, -0.0087, -0.0045, ..., -0.0044, -0.0139, -0.0038], + dtype=torch.bfloat16)} +[06-26 08:28:40|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:40|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 08:28:40|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 0 checkpoint. +[06-26 08:28:40|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/blur(ba2f44f2-c726-4fe7-949f-597069d9b91c) +[06-26 08:28:40|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 1 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/eb5325b77d358944da58a690157dd2b8071bbf85/general/blur/ba2f44f2-c726-4fe7-949f-597069d9b91c_ema_bf16.pt +[06-26 08:28:41|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0142, 0.0030, 0.0065], + [-0.0017, 0.0060, -0.0327, ..., -0.0070, 0.0009, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0135, -0.0033, 0.0283], + ..., + [-0.0031, 0.0125, 0.0248, ..., 0.0437, -0.0069, -0.0192], + [ 0.0277, 0.0146, 0.0040, ..., -0.0038, -0.0090, 0.0146], + [ 0.0437, 0.0366, 0.0085, ..., 0.0038, -0.0129, 0.0116]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0045, -0.0310, 0.0300, ..., -0.0068, -0.0048, -0.0339], + [-0.0103, -0.0048, 0.0115, ..., 0.0048, 0.0092, 0.0520], + [-0.0061, 0.0132, 0.0398, ..., -0.0096, -0.0179, -0.0061], + ..., + [-0.0046, 0.0175, -0.0095, ..., 0.0225, 0.0010, 0.0030], + [-0.0209, -0.0081, 0.0087, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0055, 0.0354, -0.0081, ..., 0.0118, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0011, ..., 0.0376, 0.0532, 0.0014], + [ 0.0175, -0.0347, -0.0013, ..., -0.0057, 0.0050, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0154, 0.0262, -0.0112], + [-0.0001, 0.0244, 0.0356, ..., 0.0019, 0.0170, 0.0119], + [ 0.0486, 0.0175, 0.0042, ..., -0.0020, -0.0023, -0.0021]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0141, 0.0569, 0.0115, ..., -0.0063, -0.0337, 0.0076], + [ 0.0491, 0.0413, 0.0251, ..., -0.0053, 0.0393, 0.0047], + [-0.0162, -0.0033, 0.0134, ..., -0.0198, -0.0244, 0.0469], + ..., + [ 0.0014, 0.0261, 0.0022, ..., -0.0041, 0.0347, -0.0033], + [-0.0254, 0.0068, -0.0016, ..., 0.0096, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0122, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0033, -0.0270, -0.0166, ..., 0.0118, -0.0298, 0.0532], + [-0.0031, -0.0035, -0.0128, ..., -0.0161, 0.0149, 0.0308], + [ 0.0144, 0.0086, 0.0146, ..., -0.0118, -0.0410, 0.0354], + ..., + [ 0.0062, -0.0630, -0.0034, ..., 0.0072, -0.0183, 0.0217], + [ 0.0038, 0.0036, 0.0258, ..., -0.0486, -0.0349, -0.0036], + [ 0.0182, -0.0081, 0.0197, ..., 0.0150, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0044, -0.0028, -0.0005, ..., 0.0003, 0.0015, -0.0123], + [-0.0150, 0.0251, 0.0432, ..., -0.0049, 0.0038, -0.0052], + [-0.0153, -0.0203, -0.0245, ..., -0.0120, -0.0325, 0.0040], + ..., + [ 0.0026, 0.0023, 0.0417, ..., 0.0280, 0.0030, -0.0042], + [-0.0024, 0.0001, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0065, -0.0019, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0155, -0.0062, -0.0481, ..., 0.0449, 0.0212, 0.0129], + [ 0.0154, -0.0227, 0.0220, ..., -0.0221, 0.0036, 0.0080], + [-0.0151, -0.0099, 0.0430, ..., -0.0095, 0.0327, -0.0111], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0139], + [-0.0017, 0.0166, -0.0059, ..., 0.0027, 0.0080, -0.0071], + [-0.0110, -0.0028, 0.0117, ..., 0.0058, -0.0034, -0.0075]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0040, -0.0386, -0.0118, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0079, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0126, -0.0044, -0.0126, ..., 0.0216, 0.0103, -0.0055], + ..., + [-0.0152, -0.0172, -0.0137, ..., 0.0098, 0.0098, 0.0184], + [-0.0152, -0.0356, 0.0201, ..., -0.0055, -0.0115, 0.0332], + [-0.0396, -0.0123, -0.0090, ..., -0.0284, -0.0142, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0277, 0.0312, 0.0133, ..., -0.0195, 0.0116, -0.0146], + [ 0.0192, 0.0095, -0.0075, ..., 0.0027, -0.0129, -0.0198], + [-0.0251, 0.0114, -0.0137, ..., 0.0093, 0.0172, 0.0075], + ..., + [ 0.0466, 0.0037, -0.0007, ..., -0.0044, 0.0239, 0.0031], + [ 0.0547, -0.0194, -0.0108, ..., -0.0056, 0.0311, -0.0023], + [ 0.0048, 0.0099, -0.0096, ..., 0.0203, 0.0170, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0045, -0.0032, 0.0283, ..., -0.0110, 0.0031, -0.0182], + [ 0.0049, 0.0129, -0.0315, ..., -0.0457, -0.0063, 0.0106], + [ 0.0027, 0.0261, 0.0245, ..., 0.0046, 0.0080, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0154, 0.0045, -0.0075], + [ 0.0055, -0.0371, 0.0356, ..., -0.0081, 0.0410, -0.0178], + [ 0.0178, -0.0151, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0079, 0.0208, 0.0273, ..., 0.0136, 0.0156, 0.0073], + [-0.0074, 0.0023, 0.0038, ..., 0.0146, 0.0015, 0.0050], + [ 0.0045, -0.0005, 0.0244, ..., -0.0120, -0.0055, -0.0247], + ..., + [ 0.0245, 0.0211, -0.0069, ..., 0.0247, -0.0092, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., -0.0052, -0.0295, 0.0400], + [-0.0116, 0.0132, 0.0305, ..., -0.0108, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0072, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0150, -0.0043, 0.0288, ..., 0.0141, 0.0583, -0.0025], + [-0.0281, 0.0156, -0.0386, ..., -0.0444, -0.0093, -0.0038], + ..., + [ 0.0041, 0.0157, 0.0217, ..., 0.0019, 0.0114, 0.0153], + [ 0.0059, 0.0195, -0.0283, ..., 0.0031, -0.0010, -0.0342], + [-0.0170, -0.0195, -0.0026, ..., 0.0049, 0.0165, 0.0133]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0139, -0.0146, 0.0459, ..., 0.0044, -0.0320, 0.0277], + [ 0.0033, 0.0064, 0.0270, ..., -0.0221, -0.0183, -0.0271], + [ 0.0221, 0.0219, -0.0201, ..., 0.0043, 0.0073, -0.0046], + ..., + [-0.0258, -0.0023, 0.0215, ..., 0.0041, 0.0237, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, 0.0014], + [ 0.0042, -0.0081, -0.0129, ..., -0.0334, 0.0060, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0145, -0.0454, ..., -0.0129, -0.0128, 0.0400], + [ 0.0295, -0.0187, -0.0109, ..., -0.0283, 0.0041, -0.0186], + [ 0.0052, 0.0547, 0.0033, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0280, -0.0058, -0.0269, ..., -0.0087, -0.0287, 0.0148], + [ 0.0091, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0098, 0.0104, ..., -0.0072, 0.0703, -0.0043]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0090, -0.0315, ..., 0.0078, 0.0254, 0.0359], + [ 0.0092, -0.0015, 0.0004, ..., -0.0123, -0.0233, -0.0046], + [-0.0025, 0.0046, -0.0126, ..., -0.0124, -0.0062, 0.0083], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0023], + [ 0.0339, -0.0072, 0.0223, ..., 0.0410, 0.0004, 0.0136], + [-0.0034, -0.0439, -0.0486, ..., 0.0139, -0.0129, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[-0.0035, 0.0141, -0.0068, ..., -0.0052, -0.0043, -0.0229], + [-0.0287, 0.0231, 0.0079, ..., -0.0142, -0.0036, 0.0222], + [ 0.0071, 0.0297, -0.0189, ..., -0.0168, 0.0311, -0.0262], + ..., + [ 0.0233, -0.0049, -0.0050, ..., 0.0045, 0.0278, -0.0034], + [-0.0103, 0.0035, 0.0142, ..., -0.0175, 0.0444, 0.0425], + [ 0.0078, -0.0079, 0.0099, ..., 0.0061, -0.0344, 0.0229]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[-0.0027, 0.0123, 0.0190, ..., -0.0096, -0.0198, 0.0148], + [-0.0581, 0.0294, 0.0032, ..., -0.0240, 0.0071, 0.0302], + [ 0.0087, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0040, -0.0289, 0.0537, ..., 0.0051, -0.0061, -0.0098], + [-0.0041, -0.0081, 0.0209, ..., 0.0049, 0.0101, 0.0186], + [-0.0087, -0.0349, -0.0173, ..., -0.0400, 0.0037, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0181, -0.0265, -0.0108, ..., 0.0096, -0.0133, -0.0405], + [ 0.0078, -0.0090, 0.0266, ..., -0.0400, 0.0073, 0.0391], + [-0.0045, 0.0359, 0.0284, ..., 0.0297, -0.0232, 0.0212], + ..., + [-0.0045, 0.0124, 0.0132, ..., -0.0134, 0.0016, 0.0095], + [ 0.0298, -0.0167, -0.0038, ..., -0.0070, 0.0048, -0.0206], + [ 0.0050, -0.0056, -0.0040, ..., 0.0009, -0.0044, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, 0.0002, 0.0123, ..., 0.0014, -0.0208, -0.0079], + [-0.0096, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0229], + [ 0.0137, -0.0278, 0.0117, ..., 0.0396, -0.0085, -0.0078], + ..., + [-0.0033, 0.0037, -0.0311, ..., 0.0344, 0.0062, -0.0124], + [-0.0236, -0.0100, -0.0067, ..., 0.0332, -0.0117, -0.0173], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0138, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[ 0.0038, 0.0166, 0.0041, ..., 0.0273, -0.0352, 0.0149], + [-0.0510, -0.0076, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0076, ..., -0.0388, 0.0004, 0.0398], + ..., + [-0.0160, -0.0195, -0.0052, ..., -0.0007, -0.0054, -0.0038], + [-0.0187, 0.0040, 0.0094, ..., 0.0287, -0.0122, -0.0098], + [-0.0181, -0.0038, 0.0247, ..., -0.0146, 0.0068, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0059, 0.0097, 0.0140, ..., -0.0112, -0.0019, -0.0253], + [-0.0166, 0.0330, -0.0171, ..., -0.0034, -0.0403, -0.0110], + [ 0.0084, 0.0347, 0.0275, ..., 0.0150, 0.0067, 0.0080], + ..., + [ 0.0271, -0.0167, -0.0106, ..., -0.0098, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0052, ..., 0.0033, 0.0017, -0.0243], + [-0.0079, 0.0581, 0.0067, ..., 0.0147, 0.0239, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0123, -0.0410, 0.0126, ..., -0.0254, -0.0027, -0.0173], + [ 0.0476, 0.0315, 0.0150, ..., -0.0146, -0.0070, 0.0038], + [-0.0092, -0.0270, 0.0086, ..., 0.0332, 0.0148, -0.0075], + ..., + [-0.0067, 0.0028, 0.0162, ..., -0.0177, 0.0171, -0.0134], + [ 0.0048, -0.0187, -0.0052, ..., 0.0248, -0.0140, -0.0058], + [-0.0142, 0.0022, 0.0150, ..., 0.0029, -0.0024, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0063, 0.0093, ..., 0.0255, -0.0253, -0.0132], + [-0.0048, -0.0135, -0.0160, ..., -0.0107, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0172, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0119, 0.0037, 0.0322, ..., -0.0183, -0.0033, 0.0031], + [ 0.0043, 0.0008, 0.0162, ..., 0.0028, 0.0255, 0.0071], + [ 0.0270, 0.0085, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0066, ..., -0.0532, 0.0104, -0.0042], + [-0.0240, -0.0215, -0.0103, ..., -0.0060, 0.0098, -0.0168], + [ 0.0062, -0.0046, -0.0064, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0059, 0.0172, -0.0106, ..., -0.0300, -0.0027, 0.0435], + [ 0.0049, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0066], + [-0.0167, 0.0123, 0.0101, ..., -0.0247, 0.0119, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0072, -0.0079, 0.0317, ..., 0.0199, -0.0105, -0.0376], + [ 0.0356, 0.0378, 0.0134, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., -0.0002, 0.0059, -0.0322], + ..., + [ 0.0059, 0.0107, -0.0040, ..., 0.0366, -0.0206, 0.0374], + [ 0.0024, -0.0079, 0.0040, ..., -0.0099, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0039, ..., -0.0166, -0.0059, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0048, -0.0087, -0.0344, ..., 0.0047, 0.0347, -0.0178], + [ 0.0267, -0.0366, -0.0089, ..., 0.0217, 0.0349, -0.0028], + [-0.0094, -0.0286, 0.0042, ..., -0.0026, -0.0282, 0.0029], + ..., + [-0.0128, 0.0037, 0.0053, ..., -0.0084, 0.0165, 0.0075], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0143], + [-0.0139, 0.0034, -0.0170, ..., -0.0128, 0.0061, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0028, 0.0034, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0118, 0.0374, ..., -0.0153, -0.0188, -0.0136], + ..., + [ 0.0262, -0.0019, -0.0075, ..., -0.0011, 0.0038, -0.0165], + [-0.0222, -0.0154, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0132, -0.0148, 0.0029, ..., 0.0020, -0.0234, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0102], + [ 0.0249, 0.0205, 0.0072, ..., -0.0032, -0.0092, -0.0103], + [-0.0220, -0.0022, -0.0023, ..., 0.0306, 0.0381, -0.0015], + ..., + [ 0.0200, -0.0011, 0.0170, ..., 0.0115, 0.0320, -0.0157], + [ 0.0349, 0.0117, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0067, 0.0256, -0.0023, ..., 0.0205, -0.0045, -0.0106]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0039, 0.0199, ..., 0.0013, 0.0276, -0.0312], + [ 0.0178, 0.0033, 0.0099, ..., 0.0110, -0.0398, 0.0293], + [ 0.0430, -0.0128, 0.0062, ..., 0.0040, -0.0084, 0.0067], + ..., + [ 0.0192, -0.0111, -0.0108, ..., -0.0045, 0.0330, 0.0140], + [-0.0126, -0.0542, -0.0205, ..., 0.0017, -0.0159, 0.0320], + [ 0.0303, 0.0066, -0.0425, ..., -0.0037, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0065, -0.0315, ..., 0.0525, 0.0039, -0.0349], + [-0.0019, -0.0039, -0.0154, ..., 0.0053, -0.0240, 0.0305], + [-0.0256, -0.0028, 0.0072, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0069, 0.0231, 0.0209, ..., 0.0075, 0.0045, 0.0461], + [-0.0072, -0.0079, -0.0001, ..., -0.0217, 0.0347, 0.0039], + [ 0.0032, 0.0234, 0.0038, ..., 0.0261, -0.0033, 0.0100]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0045, 0.0035], + [-0.0183, -0.0211, -0.0037, ..., -0.0145, 0.0138, -0.0162], + [-0.0066, -0.0160, 0.0070, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0096, 0.0078, ..., 0.0171, 0.0040, -0.0053], + [-0.0141, 0.0260, -0.0084, ..., 0.0078, 0.0103, -0.0166], + [ 0.0093, -0.0142, -0.0018, ..., 0.0053, 0.0056, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0070, 0.0141, -0.0053, ..., -0.0278, 0.0240, -0.0115], + [-0.0036, 0.0188, -0.0160, ..., -0.0317, -0.0043, 0.0181], + [-0.0006, 0.0310, -0.0085, ..., -0.0226, 0.0038, 0.0334], + ..., + [-0.0079, 0.0186, -0.0056, ..., 0.0228, 0.0052, 0.0046], + [ 0.0271, 0.0119, 0.0354, ..., 0.0019, 0.0198, 0.0143], + [ 0.0282, 0.0171, 0.0317, ..., -0.0110, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0114, -0.0265, -0.0110, ..., 0.0049, 0.0037, 0.0134], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0081], + [-0.0076, -0.0190, -0.0034, ..., 0.0166, -0.0135, 0.0137], + ..., + [-0.0200, 0.0018, 0.0060, ..., -0.0339, 0.0361, -0.0032], + [-0.0046, -0.0250, 0.0154, ..., -0.0022, -0.0162, 0.0220], + [-0.0044, 0.0038, 0.0205, ..., -0.0064, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0020, 0.0240, -0.0217, ..., 0.0172, -0.0028, 0.0115], + [-0.0386, 0.0103, 0.0074, ..., -0.0041, 0.0322, 0.0087], + [-0.0427, 0.0025, -0.0056, ..., 0.0068, 0.0110, -0.0139], + ..., + [-0.0084, -0.0043, -0.0264, ..., -0.0586, -0.0630, -0.0131], + [ 0.0064, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0106, -0.0610, 0.0162, ..., -0.0291, -0.0166, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0118, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0112], + [ 0.0108, -0.0101, -0.0327, ..., 0.0447, 0.0053, 0.0043], + [-0.0023, -0.0175, 0.0070, ..., -0.0076, -0.0049, 0.0026], + ..., + [ 0.0295, -0.0049, -0.0106, ..., 0.0095, -0.0160, -0.0288], + [ 0.0082, 0.0253, -0.0177, ..., 0.0095, 0.0071, 0.0066], + [-0.0135, -0.0057, 0.0137, ..., -0.0128, 0.0277, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0132, 0.0053, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0016, 0.0234, 0.0062, ..., 0.0464, 0.0069, 0.0309], + ..., + [-0.0118, -0.0175, 0.0327, ..., 0.0132, 0.0014, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, -0.0040, -0.0444, ..., 0.0060, -0.0283, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0081, -0.0101, 0.0146, ..., -0.0742, -0.0077, 0.0161], + [-0.0066, -0.0132, -0.0062, ..., 0.0039, -0.0023, 0.0238], + [ 0.0189, 0.0221, -0.0088, ..., 0.0018, 0.0408, 0.0045], + ..., + [ 0.0046, 0.0247, 0.0396, ..., 0.0043, -0.0304, -0.0183], + [ 0.0299, -0.0112, -0.0417, ..., -0.0039, -0.0376, -0.0045], + [-0.0210, 0.0168, -0.0050, ..., 0.0045, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0044, 0.0242, -0.0042, ..., -0.0245, 0.0033, -0.0070], + [ 0.0190, -0.0121, 0.0332, ..., -0.0144, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0037, ..., -0.0030, -0.0091, 0.0182], + ..., + [-0.0430, -0.0403, -0.0083, ..., 0.0132, -0.0141, 0.0055], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0090, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0050, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0247, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0093, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0061, ..., 0.0101, -0.0223, -0.0012], + ..., + [ 0.0192, 0.0292, -0.0018, ..., 0.0176, -0.0197, 0.0019], + [ 0.0265, 0.0063, 0.0056, ..., 0.0286, 0.0021, -0.0267], + [-0.0184, -0.0088, -0.0045, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0084, -0.0159, 0.0082, ..., 0.0198, 0.0051, 0.0277], + [ 0.0172, 0.0041, -0.0041, ..., -0.0003, 0.0040, 0.0067], + [ 0.0040, 0.0184, 0.0114, ..., 0.0099, -0.0053, -0.0059], + ..., + [ 0.0042, -0.0111, 0.0045, ..., -0.0096, 0.0036, 0.0121], + [-0.0427, -0.0391, -0.0192, ..., 0.0092, -0.0238, -0.0042], + [ 0.0038, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0083, -0.0161, 0.0123], + [ 0.0021, -0.0044, 0.0312, ..., 0.0071, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0071, 0.0046, 0.0278], + ..., + [ 0.0170, -0.0047, 0.0471, ..., -0.0057, 0.0038, -0.0079], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0039, -0.0203], + [ 0.0057, 0.0064, 0.0054, ..., -0.0095, 0.0013, 0.0021]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[ 0.0057, -0.0155, 0.0284, ..., -0.0104, 0.0135, -0.0052], + [ 0.0153, 0.0332, 0.0337, ..., 0.0223, -0.0165, 0.0038], + [-0.0020, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0069, 0.0062, 0.0280, ..., -0.0063, -0.0039, 0.0140], + [-0.0048, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0091, ..., -0.0035, 0.0167, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0045, ..., -0.0238, -0.0102, 0.0119], + [-0.0250, -0.0162, -0.0023, ..., -0.0084, 0.0002, -0.0369], + [ 0.0010, 0.0141, 0.0075, ..., -0.0452, -0.0045, 0.0120], + ..., + [ 0.0147, -0.0141, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0099, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0142, 0.0034, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[-0.0069, -0.0327, -0.0190, ..., 0.0042, 0.0036, -0.0209], + [ 0.0713, 0.0151, -0.0033, ..., 0.0051, 0.0186, 0.0020], + [-0.0227, -0.0092, 0.0188, ..., -0.0122, 0.0193, 0.0075], + ..., + [ 0.0116, 0.0464, 0.0083, ..., 0.0248, 0.0168, -0.0039], + [ 0.0381, -0.0036, 0.0060, ..., 0.0046, -0.0109, 0.0061], + [ 0.0292, 0.0050, 0.0209, ..., -0.0036, 0.0234, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0042, 0.0212, ..., 0.0305, 0.0137, 0.0104], + [-0.0060, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0065], + ..., + [ 0.0084, -0.0275, 0.0225, ..., -0.0039, -0.0189, -0.0107], + [-0.0125, 0.0041, -0.0045, ..., 0.0045, 0.0040, 0.0211], + [ 0.0020, -0.0060, -0.0059, ..., -0.0109, 0.0264, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[ 0.0043, -0.0076, -0.0108, ..., 0.0186, -0.0216, 0.0049], + [ 0.0039, 0.0123, -0.0231, ..., 0.0123, -0.0112, -0.0183], + [ 0.0107, 0.0103, -0.0011, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0123, 0.0146, ..., 0.0540, 0.0415, 0.0097], + [ 0.0038, 0.0087, -0.0138, ..., 0.0055, -0.0189, -0.0237], + [-0.0019, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, 0.0066, 0.0359, ..., -0.0084, -0.0121, -0.0041], + [ 0.0190, -0.0044, 0.0298, ..., 0.0126, 0.0048, -0.0200], + [ 0.0118, -0.0110, 0.0177, ..., -0.0248, 0.0229, -0.0038], + ..., + [-0.0077, -0.0048, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0025, -0.0291, ..., 0.0206, 0.0304, -0.0085], + [-0.0077, 0.0245, -0.0008, ..., -0.0413, 0.0216, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0060, -0.0157, -0.0032, ..., 0.0047, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0072], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, 0.0049], + ..., + [ 0.0033, -0.0058, 0.0258, ..., -0.0310, 0.0092, -0.0361], + [ 0.0040, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [-0.0036, 0.0157, -0.0046, ..., 0.0052, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0070, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0105, ..., 0.0104, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0037, ..., 0.0129, -0.0047, 0.0087], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [ 0.0038, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0128, 0.0206, ..., 0.0058, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0094, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0060, 0.0066, ..., 0.0015, 0.0112, 0.0598], + [-0.0060, 0.0267, -0.0065, ..., 0.0284, -0.0055, -0.0364], + ..., + [-0.0052, 0.0002, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, -0.0036, -0.0089, ..., 0.0211, 0.0134, -0.0078], + [ 0.0102, 0.0236, -0.0206, ..., -0.0134, 0.0048, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0077, 0.0294, ..., 0.0031, 0.0270, 0.0063], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0066, -0.0168], + [ 0.0201, -0.0121, 0.0126, ..., -0.0381, -0.0247, -0.0070], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0078], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0068, -0.0466], + [ 0.0327, 0.0284, -0.0039, ..., 0.0145, 0.0139, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, 0.0039, -0.0167, ..., 0.0005, 0.0170, 0.0107], + [ 0.0038, 0.0009, 0.0295, ..., -0.0172, 0.0038, 0.0135], + [ 0.0193, 0.0236, -0.0125, ..., -0.0045, -0.0369, 0.0239], + ..., + [ 0.0154, -0.0413, -0.0123, ..., -0.0354, 0.0062, -0.0320], + [-0.0266, -0.0071, -0.0327, ..., 0.0040, 0.0049, 0.0043], + [ 0.0298, 0.0356, -0.0192, ..., 0.0055, 0.0039, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0045, 0.0043, 0.0317, ..., 0.0053, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0071, -0.0096], + [-0.0166, 0.0059, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, 0.0049, 0.0260, ..., 0.0159, 0.0054, -0.0474], + [-0.0042, 0.0080, 0.0038, ..., 0.0143, 0.0557, 0.0023], + [-0.0203, 0.0273, 0.0029, ..., -0.0330, 0.0049, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0074, ..., 0.0036, -0.0093, -0.0277], + [ 0.0049, -0.0041, -0.0038, ..., -0.0049, 0.0140, -0.0161], + [-0.0049, 0.0168, -0.0057, ..., -0.0092, 0.0124, 0.0039], + ..., + [ 0.0398, 0.0452, 0.0251, ..., -0.0076, -0.0330, -0.0043], + [-0.0620, 0.0033, 0.0371, ..., 0.0410, 0.0037, -0.0173], + [ 0.0126, -0.0039, -0.0270, ..., -0.0037, 0.0388, -0.0104]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, -0.0039, 0.0245], + [-0.0449, 0.0039, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0061, ..., 0.0029, 0.0288, 0.0124], + [ 0.0330, 0.0040, 0.0059, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0074, -0.0030, ..., -0.0063, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0106, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0045, 0.0029, 0.0527], + [ 0.0041, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0086], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0095, -0.0095, 0.0327], + [ 0.0142, 0.0114, 0.0061, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0126, 0.0281, ..., -0.0242, -0.0126, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0029], + [-0.0053, 0.0039, -0.0286, ..., 0.0054, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0057, ..., -0.0037, 0.0081, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0096, 0.0150], + [-0.0025, 0.0021, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0071, ..., 0.0092, -0.0114, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0070, -0.0017, ..., -0.0220, -0.0145, 0.0040], + [-0.0166, 0.0189, -0.0065, ..., 0.0096, -0.0018, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0006, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0071, ..., -0.0041, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0070, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0122, 0.0289, 0.0125, ..., 0.0134, 0.0078, -0.0103], + [ 0.0188, -0.0197, -0.0039, ..., 0.0284, -0.0342, -0.0072], + [-0.0525, -0.0215, 0.0154, ..., -0.0270, 0.0084, 0.0254], + ..., + [-0.0068, 0.0087, 0.0281, ..., 0.0129, -0.0043, -0.0045], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0038, 0.0026], + [-0.0273, -0.0747, 0.0208, ..., 0.0041, 0.0041, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0047, 0.0150, 0.0334, ..., 0.0205, -0.0273, 0.0040], + [-0.0165, 0.0111, -0.0135, ..., 0.0106, -0.0069, 0.0219], + [ 0.0053, 0.0227, -0.0056, ..., 0.0128, -0.0070, 0.0150], + ..., + [-0.0347, -0.0148, 0.0061, ..., 0.0021, -0.0021, -0.0093], + [-0.0065, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0024, -0.0500, -0.0232, ..., 0.0320, 0.0022, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0059, -0.0364], + [ 0.0220, 0.0036, -0.0114, ..., 0.0047, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, -0.0039, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0074, 0.0192, -0.0393], + [-0.0101, -0.0244, 0.0275, ..., -0.0057, 0.0044, 0.0081], + [-0.0170, -0.0427, 0.0140, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0046, -0.0378, -0.0359, ..., 0.0552, -0.0073, -0.0160], + [ 0.0017, 0.0142, 0.0134, ..., 0.0049, -0.0153, 0.0354], + [ 0.0095, -0.0038, 0.0057, ..., -0.0016, -0.0249, -0.0135], + ..., + [ 0.0078, -0.0126, -0.0059, ..., -0.0095, -0.0084, -0.0028], + [-0.0059, 0.0439, 0.0166, ..., 0.0248, 0.0132, -0.0187], + [ 0.0175, 0.0060, 0.0043, ..., 0.0410, -0.0096, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0107, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0143, -0.0031, ..., -0.0039, -0.0244, 0.0356], + [ 0.0077, 0.0165, -0.0200, ..., -0.0048, 0.0237, -0.0071], + ..., + [-0.0197, -0.0022, 0.0386, ..., 0.0181, 0.0112, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, 0.0035, -0.0374], + [-0.0084, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0047, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0136, 0.0114, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, -0.0032, -0.0040, ..., -0.0048, -0.0030, 0.0337], + ..., + [ 0.0422, 0.0046, 0.0352, ..., -0.0304, -0.0039, -0.0204], + [ 0.0302, 0.0156, 0.0103, ..., -0.0076, -0.0254, -0.0245], + [-0.0041, -0.0069, 0.0042, ..., -0.0045, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, 0.0040, 0.0266, ..., 0.0154, 0.0187, 0.0039], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0041, ..., -0.0172, -0.0091, 0.0231], + ..., + [-0.0295, -0.0137, 0.0001, ..., -0.0041, -0.0052, 0.0099], + [-0.0045, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0075], + [ 0.0040, -0.0048, -0.0259, ..., 0.0045, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[-0.0047, 0.0062, -0.0140, ..., 0.0242, -0.0040, -0.0386], + [-0.0141, 0.0064, -0.0097, ..., -0.0369, -0.0155, 0.0157], + [-0.0267, -0.0020, -0.0068, ..., -0.0127, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0019, 0.0144, ..., 0.0115, 0.0039, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., 0.0041, -0.0039, -0.0137], + [-0.0219, -0.0396, -0.0078, ..., 0.0028, 0.0041, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0036, 0.0168, 0.0194, ..., 0.0349, -0.0420, 0.0013], + [ 0.0043, 0.0273, 0.0146, ..., -0.0184, -0.0035, 0.0154], + [ 0.0145, -0.0039, -0.0078, ..., 0.0245, 0.0359, 0.0059], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0037, 0.0044, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0073, -0.0312, -0.0457], + [ 0.0056, 0.0034, -0.0057, ..., 0.0021, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0037, 0.0121, 0.0039, ..., 0.0126, 0.0018, 0.0074], + [-0.0277, -0.0030, -0.0181, ..., 0.0038, -0.0029, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0043, 0.0513, -0.0113], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0113], + [-0.0282, -0.0042, 0.0121, ..., -0.0045, -0.0017, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0032, -0.0130, ..., 0.0141, 0.0308, 0.0056], + [ 0.0120, -0.0332, -0.0320, ..., -0.0073, -0.0142, -0.0032], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, 0.0045, -0.0311, ..., 0.0120, 0.0058, 0.0082], + [ 0.0043, 0.0198, 0.0141, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0125, -0.0297, -0.0055, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0103, ..., 0.0165, -0.0280, 0.0481], + [-0.0092, 0.0043, 0.0095, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, 0.0019, ..., 0.0302, -0.0025, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0039, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0022, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0048, -0.0222, 0.0262], + ..., + [-0.0147, 0.0049, 0.0260, ..., 0.0044, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0056, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0046, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0147, -0.0249, 0.0114], + [-0.0278, -0.0078, -0.0374, ..., 0.0276, -0.0062, 0.0150], + [ 0.0088, 0.0237, -0.0156, ..., 0.0527, -0.0075, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0072, -0.0095], + [-0.0071, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0034, -0.0129, ..., 0.0039, -0.0371, 0.0043], + [ 0.0059, -0.0127, 0.0043, ..., -0.0204, 0.0039, 0.0065], + [-0.0112, -0.0195, 0.0052, ..., -0.0542, -0.0089, -0.0045], + ..., + [-0.0205, -0.0084, 0.0038, ..., -0.0199, 0.0049, 0.0036], + [-0.0220, -0.0302, -0.0103, ..., 0.0107, 0.0369, -0.0038], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0039, 0.0049, 0.0238, ..., -0.0146, -0.0320, -0.0021], + [-0.0193, 0.0364, -0.0029, ..., -0.0061, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0063, -0.0232, -0.0464], + [-0.0110, -0.0298, -0.0195, ..., 0.0112, 0.0312, 0.0152], + [-0.0070, -0.0119, -0.0195, ..., -0.0042, 0.0109, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0060, -0.0239, -0.0123, ..., -0.0040, 0.0243, -0.0396], + [-0.0100, 0.0153, 0.0061, ..., -0.0277, -0.0045, 0.0016], + ..., + [ 0.0118, 0.0053, 0.0515, ..., 0.0327, -0.0118, -0.0064], + [-0.0040, 0.0170, -0.0006, ..., -0.0066, 0.0280, -0.0210], + [ 0.0012, 0.0075, 0.0041, ..., 0.0154, -0.0061, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0061, 0.0081, -0.0074, ..., -0.0588, 0.0187, -0.0067], + [-0.0103, 0.0199, -0.0078, ..., 0.0080, -0.0332, -0.0043], + ..., + [ 0.0045, -0.0464, 0.0277, ..., -0.0142, 0.0198, -0.0075], + [ 0.0162, 0.0075, 0.0089, ..., -0.0106, 0.0044, 0.0297], + [-0.0270, 0.0132, 0.0166, ..., 0.0081, 0.0175, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0053, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, -0.0007, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, -0.0039, -0.0201, ..., 0.0330, 0.0309, -0.0046], + ..., + [-0.0037, 0.0123, 0.0153, ..., -0.0042, -0.0037, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0039, 0.0175], + [-0.0244, -0.0058, -0.0300, ..., -0.0048, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0017, -0.0058, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, -0.0039, ..., 0.0310, 0.0088, -0.0037], + [ 0.0042, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0109, -0.0237, ..., -0.0104, 0.0157, 0.0064], + [ 0.0079, 0.0205, 0.0146, ..., 0.0262, -0.0195, 0.0082], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0080]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0029, ..., -0.0053, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., 0.0013, 0.0034, -0.0093], + [ 0.0017, 0.0002, 0.0398, ..., -0.0304, -0.0030, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0035, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0040], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0047, 0.0291], + [ 0.0132, -0.0035, -0.0339, ..., 0.0066, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, 0.0036], + [-0.0166, -0.0037, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0062, -0.0039, ..., 0.0240, -0.0037, -0.0031]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0007], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [ 0.0010, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0060, -0.0094], + [ 0.0059, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0078], + [-0.0206, 0.0417, -0.0125, ..., -0.0071, 0.0041, -0.0039], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0041, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, 0.0042, ..., -0.0164, 0.0036, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0028, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0054, -0.0231, -0.0255, ..., -0.0045, 0.0295, -0.0063], + [-0.0255, 0.0513, 0.0039, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0033, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0077], + [-0.0065, -0.0061, -0.0364, ..., -0.0175, 0.0306, -0.0077], + [-0.0051, -0.0195, 0.0019, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0075, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0067, -0.0193, -0.0140], + [-0.0166, -0.0048, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0128, -0.0076, 0.0140, ..., -0.0073, 0.0211, -0.0275], + [-0.0039, -0.0077, -0.0234, ..., 0.0087, -0.0070, -0.0166], + [ 0.0225, -0.0251, 0.0028, ..., 0.0078, 0.0236, 0.0081], + ..., + [ 0.0186, 0.0144, -0.0053, ..., 0.0272, -0.0084, -0.0092], + [-0.0040, -0.0030, 0.0071, ..., -0.0260, 0.0267, 0.0147], + [ 0.0101, 0.0009, 0.0212, ..., 0.0195, -0.0157, 0.0121]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0087, -0.0270, 0.0253, ..., 0.0099, 0.0048, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.1.proj.1.weight': tensor([[-0.0043, 0.0179, 0.0830, ..., -0.0645, 0.0269, -0.0361], + [-0.0288, -0.0048, -0.1138, ..., 0.0220, -0.0299, 0.0554], + [-0.0253, 0.1050, -0.0269, ..., -0.0189, 0.0175, -0.0332], + ..., + [ 0.0144, 0.0107, -0.0055, ..., 0.0226, -0.0234, 0.0046], + [-0.0284, -0.0649, 0.0503, ..., 0.0425, 0.0728, 0.0615], + [ 0.0162, -0.0171, 0.0017, ..., 0.0282, 0.0413, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_proj.weight': tensor([[-0.0054, -0.0018, 0.0101, ..., 0.0036, 0.0021, 0.0046], + [ 0.0051, 0.0034, -0.0007, ..., -0.0087, 0.0004, -0.0032], + [ 0.0102, -0.0041, 0.0022, ..., -0.0049, -0.0089, -0.0043], + ..., + [-0.0117, 0.0017, -0.0068, ..., 0.0032, 0.0063, -0.0047], + [ 0.0031, -0.0067, -0.0085, ..., -0.0039, -0.0055, -0.0053], + [ 0.0129, -0.0121, 0.0019, ..., 0.0051, -0.0067, 0.0019]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_norm.weight': tensor([ 1.5918e-01, -1.1572e-01, 1.1377e-01, 2.5977e-01, 7.2002e-05, + 5.5313e-04, 8.7172e-07, 7.4133e-07, 2.3041e-03, 4.2564e-10, + 1.2878e-09, -4.3165e-13, 1.2719e-12, 3.6948e-13, 4.2810e-13, + 7.7804e-13, 5.3478e-10, -1.0014e-05, -8.0327e-09, 3.6863e-17, + -1.7675e-13, 1.2612e-13, 2.4414e-02, 3.3379e-04, 2.8320e-02, + 6.4453e-01, 1.0400e-01, 4.0283e-03, 5.8838e-02, 9.5215e-02, + -1.3434e-14, -2.3761e-11, 6.0390e-10, -3.6380e-10, 8.9585e-11, + -6.4448e-07, 3.3833e-10, 2.2315e-14, 2.8848e-12, -1.6225e-09, + -5.2296e-12, 7.5033e-11, -5.2736e-15, 9.5367e-05, 2.4536e-02, + 6.1417e-04, -5.3955e-02, 5.8594e-02, 4.5117e-01, -1.0529e-03, + -1.8539e-03, 8.6799e-07, -3.5390e-08, -2.7008e-07, 2.2965e-11, + 9.8225e-10, -2.4556e-10, 1.8813e-07, 2.0227e-09, -2.2055e-11, + -2.7512e-18, -9.1073e-17, -3.8337e-16, -5.5422e-13, 2.7344e-01, + 2.4707e-01, 3.8086e-01, 5.0781e-02, -6.8545e-06, -3.0365e-03, + -6.7521e-09, 3.1710e-05, 2.0294e-03, -2.5437e-12, -1.0959e-10, + 3.0909e-13, 8.7766e-11, 1.1050e-10, 3.5749e-14, 1.7337e-12, + 1.9063e-09, 7.6834e-09, -3.8654e-11, -7.3830e-15, -3.1225e-15, + 1.7167e-11, 3.7537e-03, -1.1673e-03, 6.0156e-01, 7.9102e-02, + 5.3125e-01, 9.1553e-03, 9.8419e-04, 2.2888e-03, 1.9984e-14, + -1.2733e-09, -6.7812e-09, -9.9590e-11, -9.9681e-10, 5.9605e-06, + -3.2152e-13, 4.4409e-16, 2.5757e-14, 2.2847e-09, 2.1982e-14, + 7.9581e-11, -1.2506e-11, 1.7456e-02, -1.8234e-03, -1.6098e-03, + 4.7656e-01, 3.2227e-01, 7.0312e-02, -1.2665e-03, -1.2665e-03, + 1.6928e-05, -3.5288e-10, -7.1526e-07, -1.4097e-10, -6.1846e-10, + -5.4570e-11, 2.3865e-08, 3.2742e-11, -1.5250e-08, 5.2259e-17, + -1.6697e-17, 1.9451e-13, -1.4766e-14], dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.self_attn.k_proj.weight': tensor([[ 0.0060, 0.0034, 0.0004, ..., -0.0048, -0.0052, 0.0039], + [-0.0184, 0.0046, 0.0019, ..., 0.0067, -0.0027, 0.0020], + [ 0.0284, -0.0114, 0.0126, ..., -0.0076, 0.0044, 0.0008], + ..., + [-0.0018, 0.0026, -0.0046, ..., -0.0025, 0.0008, 0.0019], + [-0.0019, 0.0022, -0.0029, ..., -0.0026, 0.0020, 0.0021], + [-0.0021, 0.0010, -0.0025, ..., 0.0011, 0.0010, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.k_norm.weight': tensor([ 2.6758e-01, 1.9043e-01, 1.1426e-01, 2.2461e-01, 4.5538e-05, + -1.8234e-03, -8.0466e-07, -3.5018e-06, 2.4414e-03, 3.0195e-10, + 1.3461e-09, -5.6133e-13, -6.3949e-12, -1.7208e-14, 3.1619e-13, + -7.6739e-13, 6.0754e-10, 1.0133e-05, -6.8976e-09, -8.2399e-18, + 2.0517e-13, 1.1369e-13, 1.9165e-02, -6.6757e-04, 5.0293e-02, + 6.4453e-01, 1.2109e-01, -8.1177e-03, 3.9062e-03, 3.8605e-03, + -7.0083e-16, -2.1737e-10, -2.8667e-09, 3.5652e-10, 5.1841e-11, + 1.2368e-06, 3.3833e-10, -2.3093e-14, -2.9132e-12, 2.0227e-09, + -5.2864e-12, -9.0495e-11, 9.9365e-15, 1.2817e-02, 2.0264e-02, + -7.3433e-05, 1.2451e-01, 1.2061e-01, 4.7656e-01, -1.0376e-03, + 1.9684e-03, -6.7949e-06, 4.1677e-08, -5.0291e-07, 1.8616e-12, + 5.7116e-10, 1.3006e-10, -1.8626e-07, -2.0664e-09, -2.6375e-11, + -3.6863e-18, -2.9317e-16, -3.1398e-16, -5.5778e-13, 8.3496e-02, + 2.5391e-01, 3.8281e-01, -1.7383e-01, 3.1948e-05, -2.0905e-03, + -5.6578e-08, -2.1458e-06, -4.7302e-04, 1.9372e-10, 4.5929e-11, + -2.4514e-13, -7.7307e-11, 1.1914e-10, 6.5725e-14, 9.0949e-13, + -1.7026e-09, 1.0710e-08, -2.9786e-11, 2.7534e-14, 5.1903e-15, + 1.6485e-11, 2.5482e-03, -3.2043e-04, 5.9766e-01, -7.2266e-02, + 5.3125e-01, -4.3640e-03, 5.8105e-02, 9.4238e-02, -2.6201e-14, + 1.2005e-09, 5.5879e-09, 8.9130e-11, 1.0332e-09, -5.3644e-06, + -3.5882e-13, 1.4988e-15, -2.4691e-13, -2.3138e-09, -5.6399e-14, + -7.9581e-11, -1.2449e-11, 1.2390e-02, 1.6479e-02, 1.2360e-03, + 4.7656e-01, 2.9883e-01, 9.6680e-02, 9.9945e-04, -1.9264e-04, + 1.4544e-05, 3.4197e-09, 1.4668e-08, 1.3824e-10, -9.9681e-10, + -1.0823e-10, -2.7358e-08, -5.7128e-12, 1.5483e-08, 5.2692e-17, + -7.6978e-18, 1.9362e-13, -1.5721e-13], dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.self_attn.v_proj.weight': tensor([[ 0.0034, -0.0011, 0.0074, ..., -0.0056, 0.0005, -0.0013], + [ 0.0107, 0.0002, 0.0008, ..., 0.0023, -0.0015, -0.0035], + [ 0.0165, 0.0069, 0.0038, ..., -0.0015, 0.0008, 0.0028], + ..., + [ 0.0009, -0.0049, 0.0047, ..., 0.0081, -0.0025, -0.0019], + [-0.0013, -0.0104, -0.0045, ..., 0.0058, -0.0014, 0.0033], + [-0.0010, 0.0017, -0.0020, ..., -0.0033, 0.0019, -0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.output_proj.weight': tensor([[ 6.3171e-03, -2.3071e-02, -2.7222e-02, ..., -3.5858e-04, + -2.0599e-03, 8.5831e-04], + [-3.2043e-03, -8.5449e-03, 1.1353e-02, ..., 2.3499e-03, + 1.0925e-02, -5.2185e-03], + [ 4.8256e-04, 3.1128e-03, -2.5024e-03, ..., -8.8120e-04, + -5.2490e-03, 2.2888e-03], + ..., + [ 3.9673e-03, 5.1270e-03, 7.6294e-03, ..., 1.1444e-03, + -4.1809e-03, -7.2002e-05], + [-1.2085e-02, 7.8125e-03, 8.2397e-03, ..., -2.8687e-03, + 3.0212e-03, 1.2024e-02], + [-6.0425e-03, 4.6997e-03, -9.2773e-03, ..., -1.0071e-02, + 3.3528e-06, 5.1880e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_proj.weight': tensor([[-0.0041, -0.0047, -0.0041, ..., 0.0022, -0.0011, -0.0018], + [-0.0050, -0.0005, 0.0062, ..., 0.0010, -0.0060, -0.0040], + [ 0.0042, 0.0014, 0.0048, ..., -0.0018, 0.0037, 0.0048], + ..., + [-0.0006, 0.0045, -0.0013, ..., 0.0008, -0.0028, 0.0034], + [-0.0029, -0.0054, -0.0033, ..., -0.0016, -0.0026, -0.0128], + [ 0.0023, -0.0041, 0.0075, ..., -0.0037, 0.0009, 0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_norm.weight': tensor([ 1.7700e-03, 1.3447e-04, 1.0885e-08, 1.7319e-03, -2.4319e-05, + 2.3346e-03, -4.5896e-06, 4.3030e-03, 4.0283e-02, 2.7466e-02, + 3.2663e-05, 3.4424e-02, 5.6641e-01, 3.1586e-03, 2.4902e-02, + 1.1921e-04, 4.3750e-01, -1.3947e-05, 2.3041e-03, 5.0659e-03, + 1.1536e-02, 7.9727e-04, 9.2773e-02, 2.3041e-03, 1.0925e-02, + 4.1504e-03, 2.8687e-02, -1.3161e-04, 1.2268e-02, 1.9989e-03, + 3.6133e-01, 8.1787e-03, 2.6226e-05, 4.0820e-01, 5.3516e-01, + 3.3722e-03, 3.8910e-04, 3.8818e-02, 1.6895e-01, 3.1586e-03, + 3.3203e-02, 4.9133e-03, -4.6194e-07, 1.0010e-02, 4.0283e-03, + 3.8574e-02, 1.3794e-02, 1.0620e-02, 3.7689e-03, 9.6798e-05, + 9.5215e-03, 9.5963e-06, 1.9372e-06, 1.7262e-04, 2.2411e-04, + 2.5482e-03, 1.2207e-02, 8.6670e-03, 3.0518e-03, 8.3618e-03, + 5.8365e-04, -1.2779e-04, 1.0925e-02, 9.2163e-03, 2.0905e-03, + 1.2970e-03, 2.8992e-04, 2.6758e-01, 7.8125e-03, 1.6968e-02, + 3.9844e-01, 1.0586e-04, 4.3213e-02, 9.6130e-04, 3.6210e-06, + -1.0757e-03, 5.7068e-03, 8.3542e-04, 5.4932e-04, 1.3000e-02, + 3.7384e-03, 7.1106e-03, 1.8066e-01, 4.0894e-03, 1.0376e-03, + 2.4048e-02, 5.1270e-03, 4.1771e-04, 9.1171e-04, 4.7461e-01, + 4.7302e-03, 2.5330e-03, 1.4587e-02, -7.2956e-05, 2.7275e-04, + 4.2188e-01, 1.1396e-04, 2.2736e-03, -2.8558e-10, 1.1673e-03, + 4.0430e-01, 1.1683e-04, 9.8419e-04, 2.4223e-04, 7.5684e-03, + -1.0252e-04, 1.5106e-03, -2.0409e-04, 3.5248e-03, 5.7373e-03, + 1.1826e-03, 2.3365e-05, 8.4229e-03, 3.7231e-03, 5.0049e-03, + 5.3406e-04, 1.5564e-02, 1.8845e-03, 5.1270e-03, 5.2261e-04, + 7.3433e-05, 7.9956e-03, 5.4932e-03, 2.7847e-04, 1.3123e-03, + 3.4943e-03, 2.0409e-04, 3.8147e-04], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_proj.weight': tensor([[ 1.0452e-03, 7.2861e-04, 2.1057e-03, ..., -8.8120e-04, + -1.5717e-03, -8.1253e-04], + [-3.4790e-03, -2.0599e-03, -4.6997e-03, ..., -2.5787e-03, + -3.6621e-03, -2.4567e-03], + [-1.2512e-03, -1.4877e-03, -2.9907e-03, ..., -2.7466e-03, + -5.3406e-03, -2.0294e-03], + ..., + [ 2.0599e-03, -2.4719e-03, 2.5177e-03, ..., 1.4114e-03, + 1.0633e-04, 2.6703e-04], + [-8.0872e-04, 5.3024e-04, -2.4872e-03, ..., -3.3264e-03, + -5.9605e-05, 1.4877e-03], + [ 4.3106e-04, 1.8997e-03, -2.5024e-03, ..., -4.2114e-03, + 4.4632e-04, 1.0452e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_norm.weight': tensor([ 1.7624e-03, 1.3351e-04, -8.2946e-10, 1.4343e-03, 2.6703e-05, + 2.3499e-03, 5.7817e-06, 4.3030e-03, 3.9307e-02, 2.7466e-02, + 3.1710e-05, 3.4668e-02, 5.7031e-01, 3.1433e-03, 2.4414e-02, + 6.9141e-05, 4.3750e-01, -1.3828e-05, 2.1210e-03, 5.0049e-03, + 1.1536e-02, 8.9264e-04, 9.2773e-02, 2.4414e-03, 1.0864e-02, + 4.0894e-03, 2.8564e-02, -1.3065e-04, 1.1597e-02, 1.9989e-03, + 3.6133e-01, 7.7820e-03, 2.6464e-05, 4.0820e-01, 5.3516e-01, + 3.2654e-03, 5.8365e-04, 3.9062e-02, 1.6992e-01, 3.2959e-03, + 3.3203e-02, 4.8828e-03, -3.3900e-07, 9.8877e-03, 3.4943e-03, + 3.8574e-02, 1.3733e-02, 1.0559e-02, 3.7689e-03, 9.6321e-05, + 9.3994e-03, 9.8348e-06, 3.9339e-06, 1.8501e-04, 1.5869e-03, + 2.5787e-03, 1.2146e-02, 8.6670e-03, 3.5095e-03, 8.3618e-03, + 3.3379e-04, -1.2779e-04, 1.0925e-02, 9.0332e-03, 2.0905e-03, + 1.3351e-03, 2.8992e-04, 2.6758e-01, 7.7820e-03, 1.6968e-02, + 3.9844e-01, 1.0538e-04, 4.1260e-02, 8.8501e-04, 3.7402e-06, + 8.8501e-04, 4.8828e-03, 7.8964e-04, 5.4932e-04, 1.2878e-02, + 3.7689e-03, 7.0801e-03, 1.8066e-01, 3.8910e-03, 9.9945e-04, + 2.3926e-02, 6.3477e-03, 4.1771e-04, 8.8120e-04, 4.7461e-01, + 4.7607e-03, 2.4872e-03, 1.4282e-02, -7.2479e-05, 2.7275e-04, + 4.2188e-01, 6.0797e-05, 2.2583e-03, -3.5288e-10, 1.1597e-03, + 4.0430e-01, 1.2207e-04, 2.1362e-03, 2.3365e-04, 7.5378e-03, + -1.0347e-04, 1.5182e-03, -2.0409e-04, 2.8381e-03, 5.7373e-03, + 1.4877e-03, 3.7670e-05, 8.4229e-03, 3.6926e-03, 5.2185e-03, + -4.1199e-04, 1.5625e-02, 2.0142e-03, 6.0730e-03, 5.5313e-04, + -1.7548e-04, 8.0566e-03, 5.4321e-03, 1.6117e-04, 4.1425e-06, + 3.5095e-03, 2.9945e-04, 2.4319e-04], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.v_proj.weight': tensor([[-1.2390e-02, 3.1586e-03, 1.1902e-02, ..., 3.7842e-03, + 4.3640e-03, -1.0925e-02], + [-2.4658e-02, 2.7120e-06, 8.6060e-03, ..., 1.5030e-03, + -5.3101e-03, 6.1989e-05], + [-1.3000e-02, 3.0670e-03, 5.8594e-03, ..., 2.9602e-03, + 1.4648e-02, -1.2146e-02], + ..., + [ 9.6436e-03, 3.1090e-04, 1.8234e-03, ..., 1.4572e-03, + 5.8899e-03, 1.9043e-02], + [-9.1553e-03, 6.5613e-04, 1.0437e-02, ..., 1.2817e-03, + -1.7090e-02, -3.9368e-03], + [-1.5381e-02, -3.2654e-03, 5.3406e-03, ..., 1.8921e-03, + -8.4229e-03, -2.1362e-02]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.output_proj.weight': tensor([[ 0.0031, -0.0066, -0.0034, ..., -0.0017, 0.0056, -0.0074], + [ 0.0012, 0.0079, -0.0082, ..., 0.0134, 0.0061, -0.0087], + [-0.0018, -0.0019, -0.0030, ..., 0.0065, -0.0018, -0.0021], + ..., + [ 0.0019, 0.0023, 0.0046, ..., 0.0103, 0.0028, -0.0043], + [ 0.0082, -0.0067, 0.0049, ..., -0.0066, -0.0098, 0.0054], + [ 0.0154, -0.0197, -0.0110, ..., 0.0051, -0.0073, 0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_img.weight': tensor([[ 0.0057, 0.0117, -0.0008, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0077, 0.0249, 0.0291], + [-0.0092, 0.0061, -0.0121, ..., -0.0342, -0.0327, 0.0064], + ..., + [ 0.0070, -0.0040, 0.0016, ..., 0.0085, 0.0286, -0.0165], + [ 0.0071, -0.0060, -0.0100, ..., -0.0098, 0.0143, -0.0309], + [ 0.0242, -0.0104, 0.0255, ..., -0.0228, 0.0039, -0.0156]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0304, ..., -0.0220, 0.0247, 0.0033], + [ 0.0030, 0.0121, 0.0405, ..., 0.0128, -0.0410, 0.0361], + [-0.0089, 0.0113, -0.0087, ..., -0.0057, -0.0038, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0034, 0.0002], + [-0.0160, 0.0047, 0.0013, ..., 0.0134, -0.0175, -0.0044], + [-0.0026, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0102]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0048, -0.0364, ..., -0.0225, -0.0142, -0.0046], + [-0.0157, 0.0058, -0.0364, ..., 0.0099, 0.0119, 0.0079], + [-0.0040, 0.0615, 0.0044, ..., -0.0205, -0.0306, -0.0114], + ..., + [ 0.0038, -0.0040, -0.0103, ..., -0.0211, 0.0014, 0.0288], + [ 0.0266, -0.0117, -0.0052, ..., -0.0420, -0.0282, 0.0048], + [ 0.0286, 0.0325, 0.0097, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.mlp.layer1.weight': tensor([[ 5.4932e-03, -5.7373e-03, 4.5776e-03, ..., -4.4250e-03, + 3.2349e-03, -4.3945e-03], + [ 4.7607e-03, -3.5858e-03, -1.3351e-03, ..., 2.7008e-03, + -3.3722e-03, -3.7079e-03], + [ 3.9062e-03, 6.5308e-03, -1.2684e-04, ..., 9.3460e-04, + 2.3041e-03, -3.6469e-03], + ..., + [-5.8289e-03, -6.7444e-03, 3.8719e-04, ..., -9.0332e-03, + 6.6528e-03, -6.7749e-03], + [-2.0294e-03, 8.3618e-03, -5.5237e-03, ..., 6.8054e-03, + 1.2451e-02, -7.4463e-03], + [ 4.2419e-03, -9.6436e-03, -1.7700e-03, ..., 5.3883e-05, + -5.3101e-03, 1.6022e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.mlp.layer2.weight': tensor([[-0.0023, 0.0052, 0.0056, ..., 0.0024, 0.0047, -0.0038], + [-0.0092, 0.0043, 0.0011, ..., -0.0048, 0.0008, 0.0156], + [ 0.0038, 0.0040, -0.0080, ..., 0.0023, -0.0068, 0.0004], + ..., + [ 0.0035, -0.0021, 0.0041, ..., 0.0055, -0.0041, 0.0023], + [-0.0184, 0.0021, 0.0183, ..., -0.0038, 0.0152, -0.0055], + [ 0.0076, -0.0065, -0.0046, ..., 0.0020, 0.0037, -0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0496, -0.0488, -0.0474, ..., 0.0031, 0.0032, 0.0032], + [ 0.0510, 0.0500, 0.0486, ..., -0.0025, -0.0026, -0.0027], + [ 0.0503, 0.0496, 0.0486, ..., -0.0027, -0.0027, -0.0028], + ..., + [ 0.0483, 0.0476, 0.0459, ..., -0.0024, -0.0025, -0.0025], + [ 0.0449, 0.0439, 0.0425, ..., -0.0022, -0.0023, -0.0022], + [ 0.0498, 0.0486, 0.0476, ..., -0.0031, -0.0032, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0046, 0.0041, 0.0043, ..., 0.0056, 0.0057, 0.0046], + [-0.0015, 0.0024, 0.0016, ..., 0.0019, 0.0019, 0.0014], + [ 0.0006, -0.0019, -0.0016, ..., 0.0003, 0.0020, -0.0006], + ..., + [-0.0371, 0.0469, 0.0457, ..., 0.0430, 0.0391, 0.0378], + [-0.0066, 0.0071, 0.0073, ..., 0.0052, 0.0022, 0.0069], + [ 0.0089, -0.0101, -0.0099, ..., -0.0090, -0.0086, -0.0090]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_cross_attn.1.weight': tensor([[-6.6223e-03, -6.2561e-03, -5.4626e-03, ..., -6.3324e-04, + -5.6076e-04, -7.3242e-04], + [-3.2715e-02, -3.1250e-02, -3.1128e-02, ..., 2.0790e-04, + 2.1458e-04, 2.2507e-04], + [ 3.8330e-02, 3.7109e-02, 3.6621e-02, ..., -1.0395e-04, + -8.5831e-05, -1.0967e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1250e-02, ..., -2.8801e-04, + -2.2316e-04, -2.5940e-04], + [ 4.7607e-02, 4.6143e-02, 4.5166e-02, ..., -1.9455e-04, + -1.3447e-04, -1.9360e-04], + [ 3.4424e-02, 3.2959e-02, 3.2715e-02, ..., -1.8358e-05, + 5.8889e-05, 9.4771e-06]], dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_cross_attn.2.weight': tensor([[-1.4526e-02, -4.3335e-03, 3.7079e-03, ..., 4.5776e-03, + 4.6082e-03, 3.4637e-03], + [ 7.0190e-04, -4.8828e-04, 2.1973e-03, ..., 5.8746e-04, + 3.2654e-03, 2.2430e-03], + [ 4.6082e-03, -3.8605e-03, 3.8605e-03, ..., 3.8452e-03, + -7.1049e-05, 3.8605e-03], + ..., + [ 6.1340e-03, -5.5847e-03, 2.4292e-02, ..., 5.5847e-03, + 2.7100e-02, 1.9165e-02], + [ 9.7046e-03, 5.7983e-03, 1.1963e-02, ..., -5.6152e-03, + 1.3046e-03, 1.2085e-02], + [-1.3245e-02, -4.3640e-03, 5.9891e-04, ..., 4.1809e-03, + -1.2817e-03, 1.7262e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_mlp.1.weight': tensor([[-0.0072, -0.0069, -0.0063, ..., 0.0005, 0.0007, 0.0007], + [ 0.0537, 0.0520, 0.0505, ..., -0.0024, -0.0027, -0.0027], + [ 0.0425, 0.0415, 0.0405, ..., -0.0035, -0.0038, -0.0037], + ..., + [ 0.0874, 0.0850, 0.0835, ..., -0.0009, -0.0011, -0.0011], + [ 0.0054, 0.0049, 0.0030, ..., -0.0005, -0.0006, -0.0005], + [ 0.0530, 0.0518, 0.0503, ..., -0.0012, -0.0012, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_mlp.2.weight': tensor([[-0.0023, 0.0044, 0.0044, ..., 0.0054, 0.0023, 0.0067], + [-0.0008, 0.0017, 0.0017, ..., 0.0037, 0.0008, -0.0004], + [ 0.0019, 0.0015, 0.0017, ..., 0.0017, -0.0018, -0.0018], + ..., + [-0.0001, 0.0586, 0.0698, ..., 0.0488, 0.0009, 0.0188], + [ 0.0099, 0.0119, 0.0101, ..., 0.0149, -0.0098, 0.0068], + [-0.0082, 0.0038, 0.0038, ..., 0.0038, 0.0081, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.before_proj.weight': tensor([[ 5.5730e-06, -2.1362e-03, 2.8534e-03, ..., -4.4556e-03, + -2.5330e-03, -1.2894e-03], + [ 3.4485e-03, 9.2163e-03, -1.2589e-03, ..., 1.5335e-03, + 3.1586e-03, -3.7994e-03], + [-3.1586e-03, -3.9816e-05, 1.6098e-03, ..., -4.7302e-04, + 4.9210e-04, -1.7624e-03], + ..., + [-5.6839e-04, -1.8387e-03, -3.8757e-03, ..., 9.8419e-04, + 1.6251e-03, -1.9836e-03], + [-8.8215e-05, 6.7749e-03, 1.5411e-03, ..., 2.6131e-04, + 6.3782e-03, 2.7771e-03], + [-6.0120e-03, -5.4932e-03, 1.7700e-03, ..., 6.0654e-04, + 1.2817e-03, -2.1057e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.before_proj.bias': tensor([-0.0082, 0.0026, 0.0023, ..., -0.0035, -0.0018, 0.0038], + dtype=torch.bfloat16), 'net.control_blocks_1.0.after_proj.weight': tensor([[-0.0084, 0.0053, 0.0074, ..., -0.0014, 0.0119, 0.0034], + [-0.0170, 0.0038, -0.0052, ..., 0.0042, -0.0137, -0.0132], + [-0.0028, 0.0058, -0.0018, ..., -0.0056, 0.0006, -0.0009], + ..., + [-0.0042, 0.0040, -0.0036, ..., -0.0186, 0.0045, 0.0073], + [ 0.0078, -0.0033, -0.0013, ..., 0.0085, -0.0168, -0.0064], + [ 0.0062, -0.0062, -0.0072, ..., -0.0015, -0.0031, 0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.after_proj.bias': tensor([ 0.0045, -0.0116, -0.0005, ..., -0.0031, 0.0062, 0.0292], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_proj.weight': tensor([[-3.4424e-02, 9.1553e-03, -9.8267e-03, ..., 1.7334e-02, + 9.5749e-04, -9.5825e-03], + [ 1.3733e-02, -1.9165e-02, 4.2725e-03, ..., -2.0874e-02, + 9.9487e-03, -1.8799e-02], + [-5.3406e-03, -3.6163e-03, -5.1575e-03, ..., -5.1880e-03, + -3.4943e-03, 1.9531e-03], + ..., + [ 2.7924e-03, 3.7384e-03, 2.3193e-03, ..., 4.3030e-03, + 4.5471e-03, 6.9141e-05], + [-3.3264e-03, -3.7231e-03, -1.6403e-03, ..., -3.5706e-03, + 4.1389e-04, -1.7242e-03], + [ 5.6458e-03, -3.7842e-03, 3.2654e-03, ..., -1.7090e-03, + -2.4872e-03, 3.5248e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_norm.weight': tensor([ 2.4512e-01, 8.0859e-01, 1.5430e-01, 1.0469e+00, 8.6719e-01, + 1.5122e-34, 1.9861e-34, -1.9410e-34, 7.5984e-35, 5.8487e-07, + 1.0909e-34, 5.4543e-35, 2.5203e-35, -1.4294e-34, -1.6476e-34, + 9.7049e-35, 1.7676e-01, 2.2888e-04, 2.3079e-04, 9.5215e-03, + 3.4142e-04, 7.1411e-03, 4.1406e-01, 1.0703e+00, 1.1406e+00, + 4.1797e-01, 5.8838e-02, 5.2734e-01, -1.7822e-02, 1.6777e-34, + -4.0324e-34, 1.9034e-34, -3.2806e-03, 9.2697e-04, 4.7684e-04, + 3.9978e-03, 5.1498e-04, -1.5422e-34, -4.9277e-35, 1.3692e-34, + 1.3316e-34, -1.8161e-08, 1.3466e-34, 1.1094e+00, 5.5078e-01, + 4.0039e-01, 3.0664e-01, 1.2422e+00, 1.0547e-01, 1.2360e-03, + 1.1780e-02, 1.9043e-01, 2.6855e-03, -6.9965e-35, -1.6175e-34, + -1.8996e-35, 1.0193e-02, 2.7195e-07, 1.2488e-34, -8.4636e-35, + 3.8556e-35, 3.8744e-35, 1.0608e-34, -2.7836e-35, 5.5078e-01, + 2.7539e-01, 6.7578e-01, 3.6328e-01, 1.3574e-01, -9.5544e-35, + -6.8009e-34, 1.5046e-34, 4.5139e-35, 3.7998e-07, 1.1661e-34, + 1.9259e-34, 2.7836e-35, 7.5608e-35, -1.4219e-34, -1.3090e-34, + -9.3994e-03, 7.6675e-04, 4.4823e-04, 1.6632e-03, 9.3079e-04, + 4.8096e-02, 1.1016e+00, 4.3750e-01, 4.9023e-01, 1.1406e+00, + 2.8516e-01, 2.0117e-01, 6.5308e-03, -5.4769e-34, 2.6933e-34, + 2.0162e-34, 2.4986e-04, -2.4033e-04, 4.7874e-04, 4.5166e-03, + 8.5068e-04, 2.6519e-35, -4.1754e-35, 1.9259e-34, 1.1962e-34, + 1.2293e-07, -1.9259e-34, 3.0859e-01, 1.0625e+00, 1.1250e+00, + 1.2031e+00, 1.7773e-01, 7.6953e-01, 9.1553e-03, -1.9684e-03, + 2.1973e-03, 3.9978e-03, 2.8438e-34, -1.3015e-34, 3.3704e-34, + 9.9945e-04, 1.3828e-05, 6.9589e-35, 3.0394e-34, -1.1285e-35, + -2.1065e-34, 2.1535e-35, 6.3947e-35], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.self_attn.k_proj.weight': tensor([[ 0.0067, -0.0266, 0.0039, ..., -0.0156, -0.0106, -0.0172], + [ 0.0134, -0.0278, 0.0018, ..., -0.0325, 0.0129, -0.0195], + [ 0.0085, 0.0120, -0.0021, ..., -0.0072, -0.0024, -0.0022], + ..., + [ 0.0030, 0.0064, -0.0003, ..., 0.0005, 0.0013, -0.0032], + [-0.0025, -0.0030, -0.0014, ..., -0.0003, 0.0021, -0.0052], + [-0.0003, -0.0024, 0.0031, ..., -0.0048, 0.0014, -0.0027]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.k_norm.weight': tensor([ 4.0234e-01, 8.4766e-01, 2.1777e-01, 1.0625e+00, 8.6719e-01, + 5.9057e-35, 6.3947e-35, 6.7332e-35, -7.4855e-35, -2.3842e-07, + -1.0833e-34, -4.2882e-35, 4.4387e-35, -1.0006e-34, -1.4520e-34, + -1.3843e-34, 1.7676e-01, 3.7766e-04, 6.3896e-05, 4.6387e-02, + 3.3760e-04, -7.2937e-03, 5.1562e-01, 1.0703e+00, 1.1484e+00, + 3.4766e-01, 1.8066e-01, 6.1719e-01, -1.9653e-02, 2.7384e-34, + -4.8524e-35, 1.4896e-34, 3.4027e-03, 9.7656e-04, -1.9169e-04, + 4.6143e-02, 5.2643e-04, -5.0781e-35, -2.4262e-35, 8.5388e-35, + -7.7865e-35, -2.6659e-08, 1.6927e-34, 1.1094e+00, 4.1602e-01, + 1.2012e-01, -2.0605e-01, 1.2266e+00, 1.1377e-01, 1.6479e-03, + 9.2163e-03, 1.2354e-01, 2.1057e-03, 1.0457e-34, 2.0463e-34, + 4.3634e-35, 1.0132e-02, -3.3760e-08, -1.5723e-34, -7.8241e-35, + -1.2413e-35, -6.1314e-35, -1.4294e-34, -2.2569e-35, 4.9023e-01, + 3.3984e-01, 6.7969e-01, 2.5000e-01, 6.2500e-02, 1.5122e-34, + -1.0713e-33, 2.3623e-34, 2.3510e-35, -7.1526e-07, 2.4225e-34, + 2.3623e-34, 1.9184e-35, 8.0874e-35, 2.1535e-35, 1.0382e-34, + 1.7578e-02, 8.2779e-04, 1.6332e-05, 1.4267e-03, 9.3079e-04, + 4.7852e-02, 1.0938e+00, 3.0273e-01, 3.2617e-01, 1.1484e+00, + 2.3926e-01, 1.9434e-01, -2.1362e-04, -1.4520e-34, 2.5127e-34, + -2.9792e-34, -7.9155e-05, 4.1580e-04, 1.8978e-04, 4.5410e-02, + 8.5068e-04, 3.7428e-35, -4.5515e-35, -1.9259e-34, 8.5388e-35, + -1.2480e-07, -2.1516e-34, 3.5156e-01, 1.0781e+00, 1.1328e+00, + 1.2031e+00, -1.1182e-01, 7.8516e-01, 9.2163e-03, 6.6833e-03, + -1.6016e-01, -3.6011e-03, 4.3634e-34, -1.0608e-34, 4.9051e-34, + -1.6403e-03, 1.3828e-05, 2.9716e-35, 1.9184e-34, -3.7616e-35, + -1.0382e-34, -8.4259e-35, -4.7208e-35], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.self_attn.v_proj.weight': tensor([[ 0.0094, 0.0021, -0.0027, ..., -0.0101, -0.0006, 0.0021], + [ 0.0112, -0.0018, 0.0096, ..., 0.0132, -0.0034, -0.0018], + [ 0.0030, 0.0090, -0.0054, ..., 0.0119, 0.0067, -0.0090], + ..., + [ 0.0101, 0.0050, 0.0096, ..., 0.0025, -0.0022, -0.0150], + [ 0.0035, 0.0030, 0.0087, ..., 0.0302, -0.0042, 0.0026], + [-0.0038, -0.0031, -0.0037, ..., -0.0071, 0.0027, 0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.output_proj.weight': tensor([[-0.0118, -0.0044, -0.0154, ..., 0.0201, -0.0093, -0.0011], + [ 0.0154, 0.0047, 0.0190, ..., 0.0036, -0.0079, 0.0216], + [-0.0004, 0.0034, 0.0029, ..., -0.0008, 0.0037, 0.0030], + ..., + [-0.0063, 0.0133, 0.0054, ..., -0.0106, -0.0228, 0.0036], + [ 0.0079, -0.0027, -0.0053, ..., -0.0088, -0.0022, 0.0038], + [ 0.0028, 0.0053, -0.0009, ..., 0.0108, 0.0050, 0.0019]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_proj.weight': tensor([[-0.0126, 0.0129, 0.0119, ..., 0.0013, 0.0060, -0.0038], + [ 0.0101, -0.0017, 0.0041, ..., 0.0045, -0.0068, -0.0172], + [ 0.0040, 0.0110, -0.0029, ..., 0.0031, 0.0066, -0.0067], + ..., + [ 0.0084, -0.0129, 0.0032, ..., -0.0044, -0.0045, 0.0088], + [-0.0099, 0.0049, 0.0044, ..., 0.0073, 0.0020, 0.0093], + [ 0.0398, -0.0062, -0.0023, ..., 0.0098, 0.0096, -0.0177]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_norm.weight': tensor([ 2.2217e-02, 4.4922e-01, 2.2583e-02, 9.9487e-03, 4.8218e-03, + 8.4639e-06, 7.0572e-04, 5.1562e-01, 7.6675e-04, 3.9062e-01, + 2.9785e-02, 1.9653e-02, 8.4229e-03, 1.5015e-02, 1.8921e-02, + 1.4648e-02, 8.5449e-03, -9.6798e-05, 6.1512e-05, 3.1738e-02, + 2.6172e-01, 2.5635e-02, 5.4016e-03, 1.2512e-02, 6.2561e-03, + 1.2302e-04, -2.0266e-05, 4.8584e-02, 7.5102e-06, 3.7109e-01, + 2.0790e-04, 7.8082e-06, 9.6436e-03, -3.2336e-06, 1.0559e-02, + 7.2021e-03, 6.2466e-05, 3.9844e-01, 1.4343e-02, 6.5613e-03, + 7.0801e-02, 6.4392e-03, 1.0986e-02, 4.5471e-03, 1.2817e-02, + 8.6060e-03, 6.3171e-03, 1.7822e-02, 2.4902e-02, 1.3062e-02, + 1.4160e-02, 4.0234e-01, 1.1169e-02, 1.9897e-02, 3.1982e-02, + 1.3351e-03, 9.6680e-02, 1.3611e-02, 2.0264e-02, 1.4954e-02, + -5.2750e-06, 1.9043e-02, 1.5625e-02, 1.3053e-05, 1.2268e-02, + 8.2397e-03, 1.4496e-04, 8.3923e-04, 1.6357e-02, 7.8125e-02, + 1.2131e-03, 1.3123e-02, 4.8828e-02, 3.9795e-02, 1.6327e-03, + 1.8433e-02, -1.6212e-05, 2.5269e-02, 2.5635e-02, 2.5269e-02, + 6.8970e-03, 1.8555e-02, 3.4424e-02, -3.5286e-05, 2.1240e-02, + 3.3569e-03, 1.5030e-03, 4.4189e-02, 6.3782e-03, 1.4267e-03, + 1.6602e-02, 2.2030e-04, 7.4768e-03, 2.0020e-02, 9.0942e-03, + 5.9128e-05, 1.7456e-02, 4.4189e-02, 1.5991e-02, 4.2236e-02, + 3.8330e-02, 2.7588e-02, 1.2878e-02, 9.3460e-05, -3.5286e-05, + 5.5847e-03, 1.1414e-02, 2.1484e-02, 1.7456e-02, 2.5269e-02, + 7.3547e-03, 6.4087e-03, 3.5400e-02, 4.9023e-01, 3.0212e-03, + 1.6724e-02, 6.8054e-03, 1.1292e-02, 7.5989e-03, 1.0547e-01, + 1.4648e-03, 1.3733e-03, -9.6560e-06, 1.2634e-02, 5.9204e-03, + 2.1851e-02, 2.7466e-02, 1.6846e-02], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_proj.weight': tensor([[-0.0032, -0.0033, -0.0042, ..., -0.0037, -0.0011, 0.0017], + [ 0.0116, -0.0035, 0.0184, ..., 0.0208, 0.0198, 0.0116], + [-0.0008, 0.0036, 0.0039, ..., 0.0021, 0.0052, -0.0018], + ..., + [-0.0070, -0.0049, -0.0031, ..., -0.0012, 0.0021, -0.0016], + [-0.0072, -0.0053, -0.0142, ..., 0.0007, 0.0086, -0.0025], + [ 0.0007, -0.0004, -0.0055, ..., -0.0012, 0.0014, -0.0043]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_norm.weight': tensor([ 2.2827e-02, 4.4922e-01, 2.2461e-02, 9.9487e-03, 4.8523e-03, + -6.1095e-06, 7.2098e-04, 5.1562e-01, 8.0109e-04, 3.9062e-01, + 3.0518e-02, 1.9653e-02, 8.4229e-03, 1.5015e-02, 1.8921e-02, + 1.4526e-02, 9.0942e-03, -9.7275e-05, 5.2214e-05, 3.1982e-02, + 2.6172e-01, 2.4902e-02, 5.4321e-03, 1.2756e-02, 6.1951e-03, + 1.2493e-04, -1.8597e-05, 4.8096e-02, 6.1989e-06, 3.7109e-01, + 2.0885e-04, 9.1791e-06, 9.3384e-03, -1.7643e-05, 1.0559e-02, + 7.2632e-03, 6.2466e-05, 3.9844e-01, 1.4404e-02, 6.5002e-03, + 6.9824e-02, 6.4087e-03, 1.0803e-02, 4.5471e-03, 1.2756e-02, + 8.5449e-03, 6.4087e-03, 1.7700e-02, 2.4780e-02, 1.3306e-02, + 1.4160e-02, 4.0234e-01, 1.1169e-02, 1.9531e-02, 3.1982e-02, + 1.3275e-03, 9.6680e-02, 1.3611e-02, 1.9897e-02, 1.5076e-02, + -7.5102e-06, 1.9043e-02, 1.5625e-02, 1.3053e-05, 1.2207e-02, + 8.1177e-03, 1.4400e-04, 8.3542e-04, 1.6357e-02, 7.8125e-02, + 1.1597e-03, 1.3123e-02, 4.9072e-02, 3.9795e-02, 1.6632e-03, + 1.8555e-02, -2.3246e-05, 2.5024e-02, 2.5757e-02, 2.5269e-02, + 6.8970e-03, 1.8433e-02, 3.3936e-02, -3.6001e-05, 2.1118e-02, + 3.2501e-03, 1.4801e-03, 4.3945e-02, 6.6528e-03, 1.7242e-03, + 1.6968e-02, 2.5558e-04, 7.4158e-03, 1.9897e-02, 8.9111e-03, + 5.9128e-05, 1.6968e-02, 4.3701e-02, 1.5991e-02, 4.1992e-02, + 3.9062e-02, 2.7588e-02, 1.3000e-02, 9.0122e-05, -3.5524e-05, + 5.5542e-03, 1.1475e-02, 2.1240e-02, 1.7456e-02, 2.5146e-02, + 7.3547e-03, 6.1646e-03, 3.4912e-02, 4.8828e-01, 3.0670e-03, + 1.6724e-02, 6.8054e-03, 1.1169e-02, 7.7820e-03, 1.0547e-01, + 1.5488e-03, 1.3199e-03, -5.1856e-06, 1.3000e-02, 5.9814e-03, + 2.1973e-02, 2.7466e-02, 1.6724e-02], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.v_proj.weight': tensor([[-0.0178, -0.0001, -0.0044, ..., 0.0012, -0.0193, -0.0247], + [-0.0055, 0.0055, 0.0051, ..., 0.0037, 0.0095, 0.0070], + [-0.0400, -0.0034, -0.0039, ..., -0.0058, -0.0315, -0.0082], + ..., + [-0.0417, 0.0030, -0.0164, ..., 0.0028, 0.0205, 0.0055], + [ 0.0082, 0.0029, -0.0052, ..., 0.0016, 0.0115, 0.0044], + [ 0.0112, 0.0037, -0.0070, ..., 0.0033, 0.0131, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.output_proj.weight': tensor([[ 0.0064, 0.0042, 0.0162, ..., -0.0038, 0.0042, 0.0040], + [-0.0072, 0.0031, -0.0103, ..., -0.0123, 0.0031, 0.0117], + [-0.0011, 0.0020, -0.0045, ..., 0.0019, 0.0011, -0.0018], + ..., + [-0.0021, -0.0033, 0.0043, ..., -0.0018, 0.0036, 0.0046], + [ 0.0087, 0.0232, 0.0022, ..., -0.0349, 0.0075, 0.0289], + [ 0.0039, 0.0022, -0.0156, ..., -0.0105, 0.0062, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_img.weight': tensor([[ 0.0079, -0.0041, 0.0430, ..., -0.0327, 0.0048, 0.0144], + [-0.0278, 0.0280, 0.0249, ..., -0.0077, -0.0378, 0.0271], + [ 0.0139, -0.0119, 0.0312, ..., -0.0072, 0.0253, -0.0111], + ..., + [ 0.0245, -0.0152, 0.0043, ..., -0.0452, 0.0216, -0.0209], + [-0.0089, 0.0076, -0.0471, ..., -0.0107, 0.0157, -0.0083], + [ 0.0074, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0027, -0.0122, -0.0108], + [ 0.0135, 0.0048, -0.0188, ..., 0.0156, 0.0074, -0.0045], + [-0.0306, -0.0021, -0.0272, ..., 0.0103, 0.0337, 0.0017], + ..., + [-0.0034, -0.0042, 0.0049, ..., 0.0208, -0.0107, 0.0065], + [ 0.0193, 0.0210, 0.0122, ..., -0.0145, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0106, -0.0132, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img.weight': tensor([[ 0.0038, -0.0043, 0.0173, ..., 0.0083, 0.0356, 0.0217], + [-0.0498, -0.0033, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0108, 0.0043, -0.0139, ..., 0.0007, 0.0269, 0.0072], + ..., + [-0.0043, 0.0126, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0302, 0.0080, -0.0162, ..., -0.0050, -0.0074, -0.0118], + [ 0.0254, 0.0130, -0.0425, ..., -0.0264, 0.0138, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.mlp.layer1.weight': tensor([[ 0.0022, -0.0042, 0.0031, ..., 0.0037, 0.0016, -0.0054], + [-0.0031, 0.0053, 0.0071, ..., -0.0023, -0.0075, 0.0022], + [ 0.0026, 0.0027, 0.0036, ..., -0.0020, -0.0065, 0.0103], + ..., + [ 0.0281, 0.0128, -0.0095, ..., 0.0020, -0.0066, 0.0111], + [-0.0449, 0.0030, -0.0078, ..., 0.0183, 0.0075, -0.0037], + [ 0.0094, -0.0079, 0.0070, ..., -0.0075, 0.0014, -0.0247]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.mlp.layer2.weight': tensor([[-0.0096, -0.0112, 0.0197, ..., 0.0036, -0.0042, -0.0118], + [-0.0065, 0.0038, -0.0004, ..., -0.0036, -0.0001, -0.0032], + [ 0.0049, 0.0027, -0.0094, ..., -0.0126, 0.0036, 0.0025], + ..., + [ 0.0063, -0.0124, 0.0052, ..., -0.0036, -0.0082, 0.0097], + [ 0.0051, 0.0172, 0.0078, ..., 0.0063, 0.0067, 0.0023], + [ 0.0024, -0.0042, 0.0013, ..., -0.0013, -0.0005, 0.0079]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0410, -0.0398, -0.0393, ..., 0.0050, 0.0053, 0.0052], + [ 0.0327, 0.0315, 0.0305, ..., -0.0039, -0.0038, -0.0039], + [-0.0449, -0.0435, -0.0430, ..., 0.0051, 0.0052, 0.0052], + ..., + [-0.0454, -0.0442, -0.0430, ..., 0.0034, 0.0035, 0.0036], + [-0.0403, -0.0396, -0.0388, ..., 0.0037, 0.0039, 0.0039], + [-0.0231, -0.0222, -0.0211, ..., 0.0014, 0.0014, 0.0015]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_self_attn.2.weight': tensor([[ 6.8359e-03, -1.9302e-03, 5.8594e-03, ..., 6.8054e-03, + 6.8970e-03, 2.0142e-03], + [-1.6174e-03, 1.6403e-03, -1.7319e-03, ..., -2.3956e-03, + -1.7853e-03, -7.0572e-04], + [-1.0864e-02, -5.0735e-04, -9.8877e-03, ..., -1.1292e-02, + -1.0803e-02, 2.0862e-05], + ..., + [-1.8433e-02, -4.3678e-04, -2.4414e-02, ..., -2.6611e-02, + -2.5757e-02, 8.3542e-04], + [-1.0986e-02, -6.4697e-03, -1.1780e-02, ..., -1.1780e-02, + -1.0376e-02, 7.3242e-03], + [ 1.7395e-03, -6.0730e-03, 2.8534e-03, ..., 2.6550e-03, + 2.0447e-03, 5.0049e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5176e-02, 5.3711e-02, 5.3223e-02, ..., -1.0872e-04, + -1.5354e-04, -1.0729e-04], + [ 4.2114e-03, 4.1504e-03, 4.1504e-03, ..., -7.5340e-05, + -1.5163e-04, -2.6703e-04], + [-1.7700e-02, -1.6968e-02, -1.6235e-02, ..., -6.6757e-04, + -6.6757e-04, -6.7902e-04], + ..., + [-2.6001e-02, -2.6001e-02, -2.5024e-02, ..., -4.2343e-04, + -3.8147e-04, -3.4332e-04], + [-2.7100e-02, -2.6733e-02, -2.5635e-02, ..., -2.1839e-04, + -1.7262e-04, -2.2602e-04], + [ 5.4199e-02, 5.3223e-02, 5.2246e-02, ..., -1.7262e-04, + -5.4836e-05, -2.0218e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_cross_attn.2.weight': tensor([[-8.9111e-03, -4.9133e-03, -1.1353e-02, ..., -6.1646e-03, + 9.2163e-03, -9.5825e-03], + [ 9.6436e-03, -1.9150e-03, -3.1128e-03, ..., 2.9755e-03, + -8.4229e-03, 1.0071e-02], + [ 5.7983e-03, 1.8215e-04, -2.6941e-05, ..., 3.1281e-03, + -4.3335e-03, 5.3101e-03], + ..., + [ 6.0547e-02, 7.0190e-03, -4.3030e-03, ..., 5.7678e-03, + -1.6113e-02, 6.0059e-02], + [ 1.0620e-02, -4.6082e-03, -8.7891e-03, ..., 4.1504e-03, + -1.5625e-02, 1.1230e-02], + [-3.8605e-03, 9.1553e-03, -1.5503e-02, ..., 1.0376e-02, + 7.3242e-03, -3.8452e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_mlp.1.weight': tensor([[-2.8687e-02, -2.8564e-02, -2.8198e-02, ..., 3.9816e-05, + -4.6194e-06, 1.8477e-05], + [ 3.6621e-02, 3.6133e-02, 3.5889e-02, ..., -3.5248e-03, + -3.5553e-03, -3.4790e-03], + [ 3.9795e-02, 3.9062e-02, 3.8574e-02, ..., -3.6621e-03, + -3.6621e-03, -3.7079e-03], + ..., + [ 4.1016e-02, 3.9795e-02, 3.9551e-02, ..., -1.8616e-03, + -1.8845e-03, -1.8234e-03], + [ 1.9653e-02, 2.0264e-02, 2.0264e-02, ..., 1.6937e-03, + 1.6708e-03, 1.5640e-03], + [ 3.4912e-02, 3.4912e-02, 3.4668e-02, ..., -2.2583e-03, + -2.3193e-03, -2.3346e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_mlp.2.weight': tensor([[-0.0033, 0.0037, -0.0026, ..., 0.0036, 0.0035, 0.0038], + [ 0.0029, -0.0019, -0.0007, ..., -0.0013, -0.0033, -0.0018], + [ 0.0055, -0.0088, -0.0107, ..., -0.0078, -0.0045, -0.0078], + ..., + [ 0.0024, -0.0120, -0.0199, ..., -0.0101, 0.0016, -0.0090], + [ 0.0038, -0.0038, -0.0087, ..., -0.0038, -0.0038, -0.0038], + [ 0.0086, -0.0053, -0.0040, ..., -0.0063, -0.0093, -0.0060]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.after_proj.weight': tensor([[-3.2227e-02, 6.6528e-03, 3.7384e-03, ..., -1.4465e-02, + 7.3242e-03, -5.8899e-03], + [ 9.6436e-03, -6.9141e-05, 6.8054e-03, ..., 1.1902e-02, + 6.5918e-03, 1.8539e-03], + [ 1.4771e-02, -7.2021e-03, -1.2085e-02, ..., -4.2419e-03, + 5.5847e-03, 4.0894e-03], + ..., + [-7.9956e-03, 6.7139e-03, 4.1504e-03, ..., 1.5198e-02, + 6.6223e-03, 6.1340e-03], + [-1.2024e-02, -2.3499e-03, 5.7983e-03, ..., 5.2795e-03, + 6.1646e-03, 6.5613e-03], + [ 5.5847e-03, -7.5684e-03, 5.6458e-03, ..., 5.3406e-03, + -6.8359e-03, -1.2894e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.after_proj.bias': tensor([-0.0041, -0.0006, 0.0020, ..., -0.0030, -0.0014, -0.0011], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_proj.weight': tensor([[ 0.0452, -0.0189, -0.0077, ..., 0.0088, -0.0152, 0.0352], + [-0.0103, -0.0201, -0.0019, ..., -0.0052, 0.0085, -0.0058], + [ 0.0259, 0.0054, -0.0006, ..., 0.0060, -0.0057, 0.0159], + ..., + [-0.0143, -0.0048, 0.0012, ..., 0.0034, -0.0030, -0.0065], + [ 0.0223, 0.0277, 0.0019, ..., 0.0176, -0.0047, 0.0086], + [-0.0100, 0.0103, 0.0188, ..., -0.0136, -0.0041, -0.0189]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_norm.weight': tensor([ 1.4844e+00, 3.2617e-01, 4.1602e-01, 3.4570e-01, 1.3750e+00, + 2.7466e-04, 1.0437e-02, 5.0962e-06, 1.9775e-02, 5.3024e-04, + -9.7789e-08, -2.5705e-07, 1.1484e+00, 3.9978e-03, -3.8369e-13, + -1.6479e-03, -1.9379e-03, -2.3499e-03, 3.0396e-02, 1.5869e-02, + 1.5820e-01, -8.8692e-05, 4.2578e-01, 1.3594e+00, 5.0000e-01, + 5.5078e-01, 3.8477e-01, 1.6699e-01, 2.5513e-02, -2.5482e-03, + 1.3611e-02, -7.2098e-04, 1.4160e-02, 5.8899e-03, -1.2512e-03, + 3.4180e-02, 1.2988e-01, 5.9891e-04, -2.2430e-03, -4.0829e-06, + -1.2589e-04, 1.5182e-03, 2.2888e-03, 1.1562e+00, 9.0332e-02, + 1.3828e+00, 5.8203e-01, 2.0117e-01, -3.6865e-02, 4.9744e-03, + -2.3651e-03, 3.0273e-02, 1.3086e-01, 5.3048e-06, -7.1526e-05, + -1.1176e-06, -2.0981e-04, -6.3419e-05, -4.4678e-02, 4.8021e-09, + 3.5156e-02, -1.7262e-04, -2.9541e-02, -2.0020e-02, 6.0156e-01, + 1.4297e+00, 1.3281e+00, 1.3984e+00, 2.9688e-01, -1.6846e-02, + 1.3245e-02, 7.3016e-06, 2.1118e-02, -1.3367e-02, 1.4808e-07, + -2.3395e-06, 5.8105e-02, 2.7618e-03, 2.0073e-13, 8.1250e-01, + 1.0625e+00, 1.3000e-02, 1.9550e-04, 3.1982e-02, 1.1230e-01, + 9.8267e-03, 1.3359e+00, 3.1250e-01, 1.3672e+00, 1.4844e+00, + 2.2266e-01, 4.8047e-01, 4.7852e-02, -2.0630e-02, -5.2185e-03, + -6.7234e-05, -9.0332e-03, 1.4832e-02, -2.1515e-03, 3.3447e-02, + 1.2158e-01, 6.1989e-05, 1.1562e+00, 6.1095e-07, -4.6015e-05, + -3.7842e-03, 3.0100e-06, 7.1484e-01, 1.4219e+00, 1.3672e+00, + 1.5781e+00, 5.2734e-01, 2.7275e-04, 5.5542e-03, 6.3171e-03, + 3.1250e-02, 1.3770e-01, 1.4365e-05, -2.9504e-06, -6.2957e-07, + -6.6757e-05, -1.0452e-03, 1.1250e+00, 4.5262e-07, 2.8442e-02, + -1.5378e-05, 7.8125e-01, 1.1016e+00], dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.self_attn.k_proj.weight': tensor([[-0.0225, 0.0005, 0.0068, ..., 0.0205, 0.0119, 0.0415], + [-0.0079, 0.0108, -0.0090, ..., -0.0039, -0.0042, -0.0036], + [-0.0143, 0.0108, -0.0109, ..., 0.0058, 0.0182, 0.0007], + ..., + [-0.0030, -0.0069, 0.0044, ..., 0.0017, 0.0044, 0.0078], + [-0.0130, -0.0059, 0.0067, ..., 0.0051, 0.0056, 0.0075], + [ 0.0010, -0.0042, 0.0038, ..., -0.0051, -0.0101, 0.0044]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.k_norm.weight': tensor([ 1.5156e+00, 7.0703e-01, 3.8867e-01, 2.8711e-01, 1.3828e+00, + -7.5989e-03, 2.1484e-02, -4.3511e-06, 2.9663e-02, -1.7471e-03, + -1.0943e-07, -2.4959e-07, 1.1484e+00, 3.7537e-03, -3.7126e-13, + -1.6602e-02, -9.8419e-04, 8.0078e-02, 3.2227e-02, -3.7384e-03, + 5.3101e-03, 9.2030e-05, 5.8594e-01, 1.2344e+00, 8.6426e-02, + 5.4297e-01, 4.4141e-01, 1.6406e-01, 4.8828e-03, -5.6763e-03, + 1.4709e-02, -7.8201e-04, 1.3306e-02, 4.9561e-02, 1.9226e-03, + -1.7014e-03, 8.1787e-03, 5.9128e-04, 4.0894e-03, 4.0829e-06, + -1.1444e-04, 1.4114e-03, 2.2583e-03, 1.2891e+00, 6.6797e-01, + 1.3125e+00, 5.8203e-01, 2.9492e-01, -4.8828e-03, 6.2012e-02, + 3.7231e-03, -6.1951e-03, 3.3722e-03, -1.6928e-05, 6.5804e-05, + 1.1697e-06, -2.0885e-04, -3.1710e-05, -9.8877e-03, -3.7835e-09, + 5.4932e-03, -1.6975e-04, -3.5645e-02, -1.4099e-02, 7.0312e-01, + 1.3359e+00, 1.3359e+00, 1.3828e+00, 2.5977e-01, 1.5259e-02, + -2.2125e-03, 8.3447e-06, -2.8687e-03, -1.3184e-02, -1.3318e-07, + 2.3246e-06, 5.6763e-03, 3.0365e-03, 3.0376e-13, 8.1250e-01, + 1.0625e+00, 6.7871e-02, 1.3733e-04, -4.4861e-03, -4.0894e-03, + 9.9487e-03, 1.2500e+00, 7.3047e-01, 1.4766e+00, 1.5312e+00, + 4.5410e-02, 4.8828e-01, -1.4160e-02, 2.1118e-02, 1.6251e-03, + -6.7234e-05, 1.3794e-02, 5.3223e-02, 1.0071e-02, -5.9204e-03, + 8.5449e-03, -3.3617e-05, 1.1562e+00, 6.2212e-07, -8.0585e-05, + -3.8300e-03, 4.7386e-06, 3.7891e-01, 1.2891e+00, 1.3672e+00, + 1.6953e+00, 5.3125e-01, -2.5024e-02, 8.5938e-02, 6.5613e-03, + 1.7334e-02, -5.0964e-03, 1.1146e-05, 1.3709e-05, 5.3644e-07, + -2.5749e-05, -1.0529e-03, 1.1250e+00, -5.0664e-07, -7.2021e-03, + 2.2411e-05, 7.8125e-01, 1.1016e+00], dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.self_attn.v_proj.weight': tensor([[-0.0078, 0.0123, -0.0081, ..., -0.0261, 0.0009, 0.0117], + [ 0.0104, -0.0057, -0.0150, ..., 0.0039, -0.0039, 0.0118], + [-0.0068, -0.0043, 0.0041, ..., 0.0190, -0.0097, -0.0103], + ..., + [ 0.0154, -0.0128, -0.0153, ..., -0.0254, -0.0073, 0.0041], + [ 0.0048, -0.0039, 0.0100, ..., 0.0019, 0.0038, -0.0142], + [-0.0115, 0.0057, -0.0239, ..., 0.0041, 0.0121, 0.0092]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.output_proj.weight': tensor([[ 0.0040, 0.0135, -0.0322, ..., 0.0049, -0.0205, 0.0128], + [ 0.0024, -0.0211, -0.0120, ..., 0.0063, 0.0154, -0.0089], + [-0.0041, 0.0023, 0.0042, ..., 0.0028, 0.0031, 0.0042], + ..., + [ 0.0118, 0.0081, -0.0229, ..., 0.0190, 0.0114, 0.0009], + [ 0.0064, 0.0043, 0.0041, ..., 0.0077, 0.0093, 0.0035], + [ 0.0062, 0.0047, -0.0041, ..., -0.0068, 0.0107, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_proj.weight': tensor([[-0.0069, 0.0082, 0.0011, ..., -0.0103, -0.0052, 0.0002], + [ 0.0091, 0.0072, 0.0038, ..., -0.0041, -0.0059, 0.0029], + [ 0.0051, 0.0062, -0.0024, ..., -0.0035, 0.0157, -0.0038], + ..., + [ 0.0027, 0.0070, 0.0029, ..., 0.0079, 0.0060, 0.0046], + [-0.0107, 0.0010, 0.0054, ..., 0.0030, -0.0099, 0.0085], + [ 0.0070, 0.0062, -0.0022, ..., -0.0051, 0.0134, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_norm.weight': tensor([3.5400e-03, 1.4551e-01, 1.3245e-02, 5.8594e-01, 2.1240e-02, 2.3315e-02, + 8.7891e-03, 4.8523e-03, 1.7456e-02, 1.2598e-01, 8.4839e-03, 5.5664e-02, + 2.7344e-01, 1.9165e-02, 1.8945e-01, 1.5320e-02, 6.7871e-02, 1.5442e-02, + 4.1504e-03, 1.8555e-02, 1.0559e-02, 1.8652e-01, 2.8711e-01, 1.9922e-01, + 2.3315e-02, 1.5918e-01, 6.6406e-02, 1.6357e-02, 4.0625e-01, 8.6670e-03, + 1.9775e-02, 5.1172e-01, 6.6895e-02, 3.4668e-02, 1.5991e-02, 1.5991e-02, + 7.5195e-02, 4.3359e-01, 5.3125e-01, 1.2451e-02, 8.6060e-03, 1.2109e-01, + 5.6152e-03, 5.2344e-01, 3.0273e-02, 5.0293e-02, 1.6235e-02, 3.8086e-01, + 7.6294e-03, 1.8555e-02, 8.2520e-02, 1.2451e-02, 1.8799e-02, 4.6484e-01, + 1.8768e-03, 2.6978e-02, 5.4932e-02, 8.0078e-02, 6.2256e-03, 1.5991e-02, + 4.8523e-03, 2.3926e-02, 1.0559e-02, 3.6133e-01, 1.6602e-02, 3.0670e-03, + 2.1118e-02, 5.9570e-02, 5.9814e-02, 1.3611e-02, 6.6528e-03, 6.0156e-01, + 3.7575e-04, 5.7031e-01, 1.3855e-02, 3.9673e-03, 1.6602e-02, 3.0664e-01, + 2.6733e-02, 3.3722e-03, 1.0986e-01, 5.2490e-03, 6.3477e-02, 9.1553e-03, + 1.9897e-02, 8.1787e-03, 1.2390e-02, 6.5430e-02, 2.0508e-01, 1.0315e-02, + 1.5564e-03, 1.8433e-02, 4.6387e-02, 2.4170e-02, 2.6245e-02, 1.5747e-02, + 1.7212e-02, 2.3193e-02, 6.3965e-02, 2.7344e-02, 4.3213e-02, 1.5198e-02, + 3.1738e-03, 6.5430e-02, 5.8594e-01, 7.6172e-02, 6.8359e-03, 4.8047e-01, + 1.3428e-02, 3.0762e-02, 1.7456e-02, 1.3000e-02, 1.4591e-04, 9.0332e-02, + 4.7852e-01, 2.4170e-02, 2.4292e-02, 5.7422e-01, 1.2158e-01, 2.6001e-02, + 1.9775e-02, 6.8750e-01, 1.5234e-01, 1.5869e-02, 1.4404e-02, 5.4932e-03, + 2.8442e-02, 1.1292e-02], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_proj.weight': tensor([[ 2.1057e-03, -3.3875e-03, 4.6692e-03, ..., 6.9427e-04, + -4.0283e-03, 4.7302e-03], + [-7.0190e-03, 2.6321e-04, -5.3711e-03, ..., 5.0964e-03, + -7.8125e-03, -5.4626e-03], + [-5.3711e-03, -1.6174e-03, -5.2490e-03, ..., 1.2451e-02, + -1.4526e-02, -3.6926e-03], + ..., + [ 6.9580e-03, 4.6082e-03, 2.5482e-03, ..., 2.9297e-03, + -1.8768e-03, -6.6223e-03], + [ 6.2866e-03, 2.9449e-03, 1.3504e-03, ..., -1.3275e-03, + 2.6398e-03, 4.4250e-03], + [ 6.9141e-05, 6.7902e-04, -1.2589e-03, ..., -6.3782e-03, + -6.1340e-03, 3.0823e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_norm.weight': tensor([ 3.5400e-03, 1.4551e-01, 1.3062e-02, 5.8594e-01, 2.0996e-02, + 2.3193e-02, 8.9722e-03, 4.7302e-03, 1.7700e-02, 1.2598e-01, + 8.4839e-03, 5.3955e-02, 2.7344e-01, 1.9287e-02, 1.8945e-01, + 1.6113e-02, 6.7383e-02, 1.5442e-02, 4.0894e-03, 1.8555e-02, + 1.0803e-02, 1.8750e-01, 2.8906e-01, 1.9922e-01, 2.3071e-02, + 1.5918e-01, 6.5430e-02, 1.5137e-02, 4.0820e-01, 8.6670e-03, + 2.0020e-02, 5.1172e-01, 6.6895e-02, 3.4424e-02, 1.6357e-02, + 1.5625e-02, 7.5195e-02, 4.3359e-01, 5.3125e-01, 1.2451e-02, + 8.5449e-03, 1.2109e-01, 5.5847e-03, 5.2344e-01, 3.0762e-02, + 5.0293e-02, 1.6479e-02, 3.8086e-01, 7.7209e-03, 1.8433e-02, + 8.2520e-02, 1.2817e-02, 1.8799e-02, 4.6289e-01, 2.1057e-03, + 2.6978e-02, 5.4688e-02, 8.0078e-02, 6.3171e-03, 1.4343e-02, + 4.7913e-03, 2.3804e-02, 1.0681e-02, 3.6133e-01, 1.6479e-02, + 3.1281e-03, 2.0630e-02, 5.9570e-02, 5.9814e-02, 1.3611e-02, + 7.0496e-03, 6.0156e-01, 1.6022e-04, 5.7031e-01, 1.4038e-02, + -4.2114e-03, 1.6602e-02, 3.0664e-01, 2.7344e-02, 3.4637e-03, + 1.1035e-01, 5.0964e-03, 6.2988e-02, 8.8501e-03, 1.9409e-02, + 8.2397e-03, 1.3000e-02, 6.5918e-02, 2.0508e-01, 1.0193e-02, + 1.7548e-03, 1.7212e-02, 4.6631e-02, 2.4170e-02, 2.6978e-02, + 1.5869e-02, 1.7334e-02, 2.3193e-02, 6.3965e-02, 2.7344e-02, + 4.2725e-02, 1.5320e-02, 3.8452e-03, 6.5918e-02, 5.8594e-01, + 7.6172e-02, 6.6223e-03, 4.8047e-01, 1.3062e-02, 3.0518e-02, + 1.8066e-02, 1.2939e-02, 1.5259e-04, 9.0332e-02, 4.7852e-01, + 2.3682e-02, 2.4292e-02, 5.7422e-01, 1.2305e-01, 2.6245e-02, + 1.9653e-02, 6.9141e-01, 1.5137e-01, 1.4648e-02, 1.3916e-02, + 5.4321e-03, 2.7954e-02, 1.1108e-02], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.v_proj.weight': tensor([[ 0.0136, -0.0051, -0.0161, ..., 0.0037, -0.0212, -0.0067], + [-0.0162, -0.0009, -0.0035, ..., -0.0041, -0.0212, 0.0166], + [-0.0165, -0.0024, 0.0033, ..., 0.0016, 0.0272, 0.0325], + ..., + [ 0.0039, -0.0062, -0.0161, ..., 0.0018, 0.0115, 0.0039], + [ 0.0231, 0.0065, 0.0222, ..., 0.0047, 0.0070, 0.0078], + [ 0.0183, 0.0043, 0.0031, ..., 0.0067, -0.0141, -0.0238]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.output_proj.weight': tensor([[ 0.0136, 0.0078, -0.0025, ..., -0.0070, 0.0131, 0.0087], + [-0.0044, -0.0047, 0.0051, ..., 0.0121, 0.0042, -0.0007], + [ 0.0028, -0.0032, 0.0048, ..., -0.0025, 0.0024, -0.0036], + ..., + [-0.0141, -0.0090, -0.0069, ..., 0.0048, -0.0078, 0.0093], + [ 0.0165, -0.0115, 0.0109, ..., -0.0265, -0.0118, -0.0236], + [ 0.0045, -0.0073, 0.0156, ..., 0.0239, 0.0071, 0.0122]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0066, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [ 0.0018, 0.0150, -0.0159, ..., 0.0040, -0.0058, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0070, -0.0072], + [-0.0337, -0.0037, 0.0052, ..., -0.0039, 0.0172, 0.0231], + [-0.0255, -0.0015, -0.0039, ..., 0.0408, 0.0430, -0.0123]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0031], + [ 0.0198, -0.0208, 0.0129, ..., 0.0088, -0.0337, -0.0051], + [ 0.0302, 0.0101, 0.0025, ..., -0.0304, 0.0199, -0.0019], + ..., + [ 0.0171, -0.0276, -0.0075, ..., 0.0231, 0.0182, -0.0045], + [ 0.0221, 0.0388, 0.0060, ..., 0.0037, -0.0488, -0.0361], + [ 0.0040, 0.0115, -0.0058, ..., 0.0037, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img.weight': tensor([[ 0.0273, 0.0038, 0.0118, ..., -0.0015, 0.0347, 0.0615], + [-0.0258, 0.0009, -0.0197, ..., 0.0139, -0.0045, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0096, 0.0132, 0.0184, ..., 0.0110, 0.0084, 0.0515], + [-0.0010, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.mlp.layer1.weight': tensor([[-0.0015, 0.0028, 0.0081, ..., 0.0044, -0.0105, -0.0049], + [-0.0078, 0.0064, 0.0078, ..., -0.0059, -0.0128, 0.0036], + [-0.0074, 0.0156, 0.0066, ..., 0.0040, -0.0019, -0.0508], + ..., + [-0.0068, 0.0077, -0.0031, ..., 0.0070, 0.0021, -0.0077], + [ 0.0013, -0.0021, -0.0081, ..., -0.0139, -0.0040, -0.0269], + [ 0.0014, 0.0190, 0.0085, ..., 0.0118, 0.0143, -0.0080]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.mlp.layer2.weight': tensor([[-0.0039, -0.0052, -0.0156, ..., -0.0110, -0.0114, 0.0038], + [ 0.0078, -0.0022, 0.0026, ..., -0.0114, -0.0106, 0.0042], + [-0.0045, 0.0028, -0.0003, ..., 0.0067, 0.0045, 0.0061], + ..., + [-0.0017, -0.0100, -0.0094, ..., 0.0044, 0.0038, 0.0044], + [-0.0137, -0.0057, 0.0025, ..., 0.0140, 0.0033, -0.0125], + [ 0.0079, 0.0063, -0.0026, ..., -0.0060, 0.0133, -0.0347]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0299, 0.0299, 0.0286, ..., -0.0041, -0.0042, -0.0043], + [-0.0284, -0.0275, -0.0271, ..., 0.0042, 0.0046, 0.0045], + [-0.0298, -0.0292, -0.0287, ..., 0.0044, 0.0048, 0.0048], + ..., + [-0.0344, -0.0332, -0.0327, ..., 0.0020, 0.0021, 0.0023], + [-0.0337, -0.0332, -0.0325, ..., 0.0020, 0.0021, 0.0020], + [-0.0156, -0.0157, -0.0159, ..., 0.0001, 0.0001, 0.0001]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0055, -0.0037, -0.0053, ..., -0.0017, -0.0019, -0.0014], + [ 0.0044, -0.0040, -0.0042, ..., -0.0033, -0.0038, -0.0022], + [ 0.0057, -0.0066, -0.0056, ..., -0.0018, -0.0021, 0.0015], + ..., + [ 0.0420, -0.0452, -0.0454, ..., -0.0278, -0.0244, 0.0075], + [-0.0060, 0.0142, 0.0075, ..., 0.0167, 0.0153, 0.0170], + [-0.0189, 0.0197, 0.0195, ..., 0.0182, 0.0177, 0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5420e-02, 5.4443e-02, 5.3467e-02, ..., -1.3733e-04, + -6.7234e-05, -2.1935e-04], + [ 5.4688e-02, 5.2979e-02, 5.1025e-02, ..., 2.7275e-04, + 3.8528e-04, 4.1771e-04], + [-4.5654e-02, -4.4434e-02, -4.3213e-02, ..., -3.8445e-06, + 5.1498e-05, -5.6744e-05], + ..., + [ 2.8809e-02, 2.7954e-02, 2.7466e-02, ..., 1.0109e-04, + 3.3140e-05, 6.2943e-05], + [ 5.9814e-02, 5.8594e-02, 5.7861e-02, ..., 4.6921e-04, + 5.1117e-04, 5.0354e-04], + [-4.7607e-02, -4.6631e-02, -4.5654e-02, ..., -2.9755e-04, + -3.4142e-04, -3.0327e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0041, 0.0135, -0.0018, ..., -0.0096, -0.0148, 0.0106], + [-0.0134, 0.0061, 0.0108, ..., -0.0055, -0.0068, 0.0040], + [-0.0006, 0.0032, 0.0011, ..., -0.0012, -0.0017, 0.0017], + ..., + [ 0.0178, 0.0061, -0.0118, ..., 0.0077, 0.0155, 0.0041], + [-0.0043, 0.0393, 0.0046, ..., -0.0013, 0.0275, -0.0136], + [-0.0101, 0.0148, 0.0110, ..., -0.0106, -0.0070, 0.0193]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_mlp.1.weight': tensor([[-0.0518, -0.0491, -0.0476, ..., 0.0030, 0.0032, 0.0032], + [-0.0063, 0.0039, 0.0039, ..., 0.0007, 0.0008, 0.0007], + [ 0.0645, 0.0625, 0.0623, ..., -0.0016, -0.0016, -0.0016], + ..., + [ 0.0588, 0.0574, 0.0566, ..., -0.0017, -0.0017, -0.0015], + [ 0.0151, 0.0139, 0.0133, ..., -0.0010, -0.0009, -0.0010], + [-0.0598, -0.0586, -0.0576, ..., 0.0021, 0.0021, 0.0020]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_mlp.2.weight': tensor([[-0.0337, -0.0108, 0.0143, ..., 0.0069, 0.0125, -0.0147], + [ 0.0018, -0.0047, -0.0014, ..., -0.0011, 0.0047, 0.0011], + [ 0.0029, -0.0035, -0.0018, ..., -0.0018, 0.0041, 0.0018], + ..., + [ 0.0403, -0.0041, -0.0137, ..., -0.0083, 0.0041, 0.0143], + [ 0.0270, 0.0325, 0.0256, ..., 0.0282, -0.0325, -0.0258], + [ 0.0095, -0.0148, -0.0086, ..., -0.0112, 0.0151, 0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.after_proj.weight': tensor([[-0.0071, -0.0042, 0.0067, ..., -0.0049, 0.0068, -0.0044], + [ 0.0003, 0.0037, 0.0046, ..., -0.0112, -0.0070, -0.0019], + [-0.0046, -0.0081, 0.0173, ..., 0.0039, 0.0036, 0.0035], + ..., + [-0.0026, 0.0045, 0.0090, ..., -0.0181, 0.0084, 0.0073], + [ 0.0036, -0.0075, -0.0081, ..., 0.0049, -0.0157, 0.0065], + [-0.0063, -0.0060, 0.0150, ..., -0.0077, 0.0002, 0.0079]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.after_proj.bias': tensor([-0.0017, -0.0015, 0.0014, ..., -0.0042, 0.0008, 0.0029], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_proj.weight': tensor([[-0.0034, 0.0080, 0.0164, ..., -0.0070, 0.0063, -0.0270], + [-0.0066, 0.0023, -0.0067, ..., -0.0078, -0.0089, -0.0082], + [ 0.0114, 0.0062, 0.0042, ..., -0.0023, -0.0038, -0.0038], + ..., + [ 0.0045, -0.0076, -0.0040, ..., -0.0042, 0.0046, 0.0069], + [-0.0044, -0.0320, -0.0098, ..., -0.0381, -0.0189, -0.0298], + [-0.0075, -0.0038, -0.0040, ..., 0.0037, -0.0047, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_norm.weight': tensor([ 1.2656e+00, 1.6562e+00, 1.5078e+00, 1.7188e+00, 1.0498e-01, + 6.9824e-02, 4.7607e-02, 1.0192e-05, -1.2757e-30, -1.5564e-03, + 2.6978e-02, 1.0156e-01, -2.2095e-02, -2.9297e-02, 9.0790e-04, + 3.2616e-04, 6.0059e-02, 9.4922e-01, 9.9609e-01, 1.4267e-03, + 6.1035e-05, 1.1953e+00, 1.6406e+00, 1.6719e+00, 6.7969e-01, + 6.2891e-01, 1.5703e+00, -3.4332e-03, -1.0071e-02, 2.0146e-05, + 4.3869e-04, -5.0354e-03, 2.3682e-02, 1.0938e+00, 5.4932e-03, + 1.0781e+00, 1.5068e-04, 4.7207e-05, 1.3123e-02, -1.6327e-03, + 8.5449e-03, -3.2187e-05, 1.2054e-03, 1.6953e+00, 9.8438e-01, + 1.4062e+00, 1.6406e+00, 1.8516e+00, 1.0596e-01, 4.9973e-04, + 3.0029e-02, 2.2168e-01, 1.0859e+00, 8.2520e-02, 2.1820e-03, + 1.3086e-01, 5.6458e-04, 9.3750e-02, 7.4863e-05, -1.1902e-02, + 1.4267e-03, 3.8300e-03, -3.3722e-03, 1.1406e+00, 1.8828e+00, + 7.9297e-01, 9.5703e-01, 1.5312e+00, 4.0771e-02, -1.2390e-02, + 1.2085e-02, -1.9908e-05, 2.5761e-30, 1.1658e-02, 1.7822e-02, + 8.9722e-03, 6.0730e-03, 4.0588e-03, -2.5269e-02, 1.8066e-02, + 5.7861e-02, -2.6245e-03, 1.4587e-02, 2.7618e-03, 3.3379e-04, + 4.1504e-03, 5.5469e-01, 4.2578e-01, 1.4688e+00, 1.6406e+00, + 4.7656e-01, 5.7068e-03, -7.9727e-04, 2.9755e-04, -4.3335e-03, + 1.0300e-03, 3.4912e-02, 6.8848e-02, 5.2795e-03, 5.2002e-02, + 2.6321e-04, 4.1485e-05, 1.2891e-01, -2.6321e-04, 6.5918e-03, + -1.4484e-05, 1.3638e-04, 4.1797e-01, 1.5000e+00, 1.4062e+00, + 5.9375e-01, 9.2969e-01, 7.3242e-02, -2.4109e-03, 2.0020e-02, + 1.1641e+00, 5.0049e-02, 1.0859e+00, -1.5736e-05, 2.3651e-03, + 3.9307e-02, 4.5410e-02, 1.4210e-04, 1.9302e-03, -1.0872e-04, + -3.6011e-03, 9.4141e-01, -2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.self_attn.k_proj.weight': tensor([[-4.2725e-02, 1.7700e-03, -2.9449e-03, ..., -3.2959e-03, + -3.9101e-05, -1.3123e-02], + [-3.5156e-02, -1.2817e-02, -1.3123e-02, ..., 1.0071e-02, + -1.1230e-02, 2.4414e-02], + [ 4.5410e-02, 4.3030e-03, 3.8910e-03, ..., -1.2146e-02, + -4.1809e-03, -2.7100e-02], + ..., + [-7.1411e-03, 1.5259e-02, 5.7602e-04, ..., -5.2795e-03, + -6.2561e-03, -2.8076e-03], + [ 1.1597e-02, -8.8501e-03, -2.4261e-03, ..., -4.0894e-03, + 1.0986e-02, 3.2349e-03], + [ 2.0294e-03, 5.9814e-03, -1.2970e-03, ..., -5.9509e-03, + -6.1035e-04, 6.5308e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.k_norm.weight': tensor([ 7.4609e-01, 1.5938e+00, 1.5234e+00, 1.6562e+00, 1.0742e-01, + 7.5195e-02, 4.6387e-02, -1.3113e-05, 8.1351e-31, -2.1210e-03, + 2.6611e-02, 1.0205e-01, -2.6245e-02, 5.2979e-02, 1.1597e-03, + 1.4114e-04, 4.0588e-03, 9.4922e-01, 9.9609e-01, 1.3275e-03, + 1.0681e-04, 1.1953e+00, 1.4531e+00, 1.5469e+00, 3.5400e-02, + 1.4355e-01, 1.5938e+00, -5.7373e-03, 9.6436e-03, -9.6798e-05, + -1.5640e-03, 5.0964e-03, -2.0630e-02, 1.0938e+00, -2.0142e-03, + 1.0703e+00, 1.6975e-04, 4.4584e-05, -1.5076e-02, -1.7395e-03, + 8.5449e-03, -7.4387e-05, 1.2054e-03, 1.4609e+00, 4.4727e-01, + 1.3750e+00, 1.6953e+00, 1.8672e+00, -5.3955e-02, 1.8387e-03, + 1.0681e-02, 1.5625e-02, 1.0859e+00, 7.6172e-02, 2.1820e-03, + 1.3086e-01, -9.0122e-05, 9.3750e-02, 7.0095e-05, 7.6660e-02, + -1.4191e-03, 1.8311e-02, -1.9653e-02, 1.1406e+00, 1.9688e+00, + 1.1172e+00, 8.6328e-01, 1.5781e+00, 2.7710e-02, -5.7983e-03, + -1.3916e-02, 1.5974e-05, 1.9352e-30, 1.2024e-02, 3.1006e-02, + 5.0781e-02, 6.3782e-03, 2.6611e-02, -3.7842e-02, -1.8066e-02, + -1.8555e-02, 6.0730e-03, 9.0332e-03, 2.6550e-03, 3.2425e-04, + 8.6670e-03, 1.1172e+00, 1.0000e+00, 1.5625e+00, 1.7031e+00, + 1.4844e-01, 5.7678e-03, -3.0212e-03, -2.6131e-04, -3.6163e-03, + -6.2943e-04, 1.9287e-02, 9.0332e-03, 4.6387e-03, -8.9722e-03, + 2.7084e-04, -1.6332e-05, 1.2793e-01, -2.6321e-04, 6.5613e-03, + 2.1100e-05, 1.3256e-04, 1.1719e+00, 1.6406e+00, 1.4141e+00, + 1.1768e-01, 9.8438e-01, 8.7280e-03, 4.0894e-03, 5.0354e-03, + 1.1641e+00, -2.9755e-03, 1.0938e+00, -7.2002e-05, -1.8158e-03, + 3.9307e-02, 4.5166e-02, 1.3924e-04, 4.3457e-02, -1.1730e-04, + 1.5015e-02, 9.4141e-01, -1.5503e-02], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.self_attn.v_proj.weight': tensor([[-0.0017, 0.0042, 0.0115, ..., 0.0542, -0.0092, 0.0073], + [ 0.0060, 0.0214, -0.0223, ..., -0.0248, -0.0010, -0.0054], + [-0.0009, 0.0156, -0.0039, ..., 0.0126, -0.0044, 0.0096], + ..., + [ 0.0026, -0.0039, -0.0148, ..., 0.0102, 0.0055, -0.0009], + [ 0.0059, 0.0243, 0.0019, ..., 0.0025, 0.0025, 0.0002], + [-0.0074, -0.0142, 0.0027, ..., 0.0175, -0.0077, -0.0128]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.output_proj.weight': tensor([[ 0.0179, 0.0205, -0.0031, ..., 0.0021, 0.0260, -0.0159], + [ 0.0096, -0.0099, -0.0095, ..., -0.0045, -0.0166, 0.0025], + [ 0.0078, 0.0236, -0.0091, ..., 0.0062, -0.0017, -0.0050], + ..., + [-0.0287, -0.0366, -0.0045, ..., -0.0045, -0.0148, 0.0040], + [ 0.0026, 0.0087, -0.0182, ..., -0.0103, 0.0327, -0.0085], + [-0.0432, -0.0049, 0.0156, ..., 0.0238, -0.0083, -0.0036]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_proj.weight': tensor([[ 0.0049, 0.0067, 0.0038, ..., -0.0072, -0.0022, 0.0223], + [ 0.0061, -0.0040, -0.0039, ..., -0.0063, 0.0070, 0.0106], + [ 0.0023, 0.0029, -0.0030, ..., 0.0027, 0.0157, 0.0051], + ..., + [ 0.0076, -0.0088, -0.0059, ..., 0.0037, -0.0072, -0.0011], + [ 0.0017, -0.0032, -0.0029, ..., 0.0056, -0.0051, -0.0022], + [-0.0040, -0.0054, 0.0039, ..., -0.0080, 0.0072, 0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_norm.weight': tensor([0.1089, 0.0232, 0.4688, 0.0752, 0.1816, 0.0356, 0.1494, 0.1260, 0.5234, + 0.0388, 0.0630, 0.0148, 0.0630, 0.0762, 0.0508, 0.2021, 0.4883, 0.0260, + 0.7383, 0.0410, 0.0311, 0.1641, 0.0289, 0.0806, 0.0557, 0.4199, 0.5156, + 0.2754, 0.0713, 0.0454, 0.0233, 0.2451, 0.7031, 0.2617, 0.0415, 0.0344, + 0.0190, 0.1025, 0.0284, 0.0422, 0.0586, 0.1318, 0.0247, 0.0344, 0.2246, + 0.0425, 0.4043, 0.0304, 0.0540, 0.0095, 0.0259, 0.4141, 0.4902, 0.3867, + 0.0962, 0.6133, 0.5781, 0.0337, 0.1494, 0.0249, 0.0898, 0.4902, 0.0215, + 0.5547, 0.1699, 0.4180, 0.1040, 0.0488, 0.0723, 0.4941, 0.0262, 0.4902, + 0.0198, 0.0447, 0.1436, 0.0850, 0.0280, 0.1128, 0.6445, 0.0308, 0.0908, + 0.0874, 0.0417, 0.0381, 0.4297, 0.0723, 0.0515, 0.0874, 0.0211, 0.1787, + 0.0398, 0.1924, 0.0403, 0.0669, 0.0442, 0.1133, 0.0052, 0.3691, 0.5234, + 0.0471, 0.0581, 0.3301, 0.0347, 0.1270, 0.0352, 0.2656, 0.0214, 0.5664, + 0.5664, 0.0391, 0.0176, 0.7500, 0.0310, 0.0139, 0.5039, 0.2217, 0.3145, + 0.7109, 0.0332, 0.0913, 0.0884, 0.5391, 0.6016, 0.0146, 0.0248, 0.2041, + 0.0415, 0.0378], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_proj.weight': tensor([[-1.7548e-03, 1.8158e-03, 3.5706e-03, ..., -3.2349e-03, + 1.1353e-02, -1.2512e-02], + [ 5.9204e-03, -6.7711e-05, -2.9602e-03, ..., -3.0518e-04, + -1.3275e-03, 6.3171e-03], + [-1.1292e-02, 7.6904e-03, -1.3245e-02, ..., 6.8970e-03, + 9.8267e-03, 1.7853e-03], + ..., + [-1.6479e-03, 3.1128e-03, -5.8899e-03, ..., -3.5095e-03, + -1.3306e-02, 9.3384e-03], + [-1.2451e-02, -1.8539e-03, -8.9111e-03, ..., 7.5684e-03, + -9.2163e-03, -2.7275e-04], + [ 1.4893e-02, -1.4038e-03, 5.0049e-03, ..., -1.0071e-02, + 4.8523e-03, -3.5400e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_norm.weight': tensor([0.1084, 0.0232, 0.4688, 0.0752, 0.1816, 0.0361, 0.1494, 0.1260, 0.5234, + 0.0393, 0.0640, 0.0149, 0.0630, 0.0762, 0.0510, 0.2041, 0.4883, 0.0262, + 0.7344, 0.0410, 0.0309, 0.1641, 0.0291, 0.0801, 0.0557, 0.4160, 0.5156, + 0.2734, 0.0713, 0.0452, 0.0242, 0.2471, 0.7031, 0.2617, 0.0422, 0.0344, + 0.0188, 0.1025, 0.0287, 0.0417, 0.0583, 0.1328, 0.0251, 0.0349, 0.2246, + 0.0425, 0.4043, 0.0306, 0.0542, 0.0092, 0.0262, 0.4141, 0.4902, 0.3867, + 0.0962, 0.6133, 0.5820, 0.0334, 0.1494, 0.0253, 0.0898, 0.4902, 0.0219, + 0.5547, 0.1689, 0.4180, 0.1040, 0.0488, 0.0723, 0.4961, 0.0261, 0.4902, + 0.0195, 0.0444, 0.1436, 0.0845, 0.0280, 0.1138, 0.6445, 0.0305, 0.0898, + 0.0874, 0.0417, 0.0383, 0.4297, 0.0728, 0.0505, 0.0869, 0.0214, 0.1807, + 0.0408, 0.1914, 0.0400, 0.0674, 0.0447, 0.1143, 0.0052, 0.3691, 0.5234, + 0.0464, 0.0588, 0.3301, 0.0347, 0.1270, 0.0352, 0.2656, 0.0211, 0.5664, + 0.5664, 0.0391, 0.0177, 0.7500, 0.0305, 0.0141, 0.5078, 0.2217, 0.3145, + 0.7070, 0.0325, 0.0898, 0.0884, 0.5391, 0.6016, 0.0145, 0.0248, 0.2041, + 0.0427, 0.0381], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.v_proj.weight': tensor([[ 0.0099, -0.0024, -0.0236, ..., -0.0018, -0.0025, -0.0165], + [-0.0064, 0.0107, 0.0040, ..., 0.0040, 0.0023, -0.0250], + [ 0.0190, 0.0034, -0.0056, ..., 0.0033, -0.0052, -0.0396], + ..., + [ 0.0025, 0.0031, 0.0190, ..., -0.0087, -0.0259, 0.0049], + [-0.0332, -0.0007, -0.0219, ..., -0.0072, 0.0103, -0.0148], + [-0.0070, -0.0017, -0.0017, ..., -0.0082, 0.0104, 0.0071]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.output_proj.weight': tensor([[-0.0128, -0.0076, -0.0082, ..., -0.0063, 0.0133, 0.0111], + [-0.0248, -0.0042, -0.0100, ..., 0.0048, -0.0138, -0.0304], + [ 0.0011, 0.0024, 0.0064, ..., 0.0029, -0.0026, -0.0045], + ..., + [ 0.0141, -0.0109, 0.0123, ..., 0.0160, -0.0127, -0.0093], + [-0.0161, -0.0079, 0.0060, ..., 0.0239, -0.0284, 0.0334], + [-0.0165, -0.0115, 0.0254, ..., -0.0051, 0.0259, -0.0051]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_img.weight': tensor([[ 1.6235e-02, -4.2480e-02, -2.7954e-02, ..., 2.3682e-02, + -4.1748e-02, -4.6631e-02], + [-1.0925e-02, 2.9297e-02, 2.5635e-02, ..., -2.3315e-02, + -1.9775e-02, -2.3193e-02], + [-1.2146e-02, 4.9438e-03, -6.5918e-03, ..., 2.6611e-02, + -2.7832e-02, 7.5989e-03], + ..., + [-2.1729e-02, -4.1016e-02, 4.8828e-02, ..., 3.8862e-05, + -2.3956e-03, -2.5513e-02], + [ 9.9182e-04, 6.6833e-03, 3.8300e-03, ..., 3.0029e-02, + 8.4229e-03, 2.0508e-02], + [-2.1362e-02, 2.1729e-02, 5.6458e-03, ..., -1.3916e-02, + -1.4282e-02, -4.1992e-02]], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, 0.0013, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0024, 0.0293, ..., 0.0067, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0139, 0.0126, -0.0092], + ..., + [-0.0032, 0.0320, -0.0233, ..., -0.0062, -0.0028, 0.0188], + [ 0.0135, -0.0352, 0.0267, ..., -0.0305, -0.0039, 0.0018], + [-0.0205, 0.0140, 0.0039, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, 0.0005, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0056, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0015], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, 0.0013, 0.0012], + [-0.0141, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0023, ..., -0.0103, 0.0071, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.mlp.layer1.weight': tensor([[ 0.0104, 0.0049, -0.0160, ..., 0.0096, 0.0089, -0.0065], + [-0.0109, -0.0065, 0.0049, ..., -0.0120, -0.0062, -0.0079], + [-0.0034, 0.0359, -0.0165, ..., 0.0284, 0.0075, 0.0089], + ..., + [-0.0074, 0.0103, 0.0645, ..., -0.0041, 0.0085, -0.0117], + [-0.0038, 0.0049, 0.0042, ..., -0.0151, 0.0087, -0.0129], + [ 0.0055, 0.0066, -0.0072, ..., -0.0146, -0.0156, -0.0186]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.mlp.layer2.weight': tensor([[ 0.0111, -0.0063, 0.0005, ..., -0.0266, -0.0067, 0.0039], + [-0.0112, 0.0051, -0.0106, ..., -0.0027, 0.0101, -0.0044], + [ 0.0006, -0.0107, 0.0055, ..., 0.0058, 0.0098, -0.0042], + ..., + [ 0.0040, 0.0040, 0.0217, ..., -0.0044, 0.0064, -0.0055], + [ 0.0053, 0.0029, 0.0039, ..., 0.0110, 0.0036, 0.0095], + [-0.0168, -0.0084, 0.0029, ..., 0.0116, 0.0139, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0413, -0.0408, -0.0405, ..., 0.0009, 0.0007, 0.0008], + [-0.0369, -0.0366, -0.0356, ..., 0.0026, 0.0027, 0.0027], + [-0.0334, -0.0332, -0.0320, ..., 0.0021, 0.0022, 0.0022], + ..., + [ 0.0244, 0.0234, 0.0225, ..., -0.0037, -0.0038, -0.0037], + [ 0.0134, 0.0123, 0.0115, ..., -0.0032, -0.0031, -0.0031], + [-0.0344, -0.0337, -0.0330, ..., 0.0022, 0.0023, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0031, 0.0044, ..., 0.0063, 0.0090, 0.0045], + [-0.0028, -0.0012, -0.0015, ..., -0.0019, -0.0014, -0.0011], + [ 0.0024, -0.0069, -0.0049, ..., 0.0032, 0.0033, -0.0058], + ..., + [-0.0237, -0.0811, -0.0635, ..., 0.0327, 0.0332, -0.0723], + [-0.0161, -0.0140, -0.0139, ..., -0.0115, -0.0139, -0.0131], + [ 0.0098, 0.0067, 0.0069, ..., 0.0092, 0.0121, 0.0064]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_cross_attn.1.weight': tensor([[-7.9102e-02, -7.6660e-02, -7.5195e-02, ..., -9.5844e-05, + -1.1110e-04, -1.0157e-04], + [ 3.0273e-02, 2.9785e-02, 2.9297e-02, ..., 2.5481e-06, + 2.8610e-05, 1.4877e-04], + [ 2.5391e-02, 2.4658e-02, 2.3682e-02, ..., 2.4796e-04, + 2.7084e-04, 1.7738e-04], + ..., + [ 5.1514e-02, 5.0293e-02, 4.9805e-02, ..., -1.8692e-04, + -1.8120e-04, -1.1349e-04], + [ 5.5176e-02, 5.3711e-02, 5.2246e-02, ..., -1.8311e-04, + -1.2207e-04, -1.7071e-04], + [-1.6235e-02, -1.5747e-02, -1.5503e-02, ..., -5.0306e-05, + 1.0109e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0053, -0.0033, ..., 0.0281, 0.0317, 0.0069], + [ 0.0040, -0.0044, 0.0056, ..., -0.0009, 0.0025, -0.0019], + [ 0.0025, -0.0038, 0.0039, ..., -0.0026, -0.0026, -0.0044], + ..., + [-0.0069, -0.0052, 0.0403, ..., 0.0016, 0.0123, -0.0041], + [-0.0302, 0.0184, 0.0298, ..., 0.0115, 0.0130, -0.0046], + [ 0.0261, -0.0206, 0.0040, ..., 0.0048, 0.0107, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_mlp.1.weight': tensor([[-7.6172e-02, -7.3730e-02, -7.2754e-02, ..., 8.7261e-05, + 2.0409e-04, 1.5354e-04], + [ 5.8105e-02, 5.5420e-02, 5.5176e-02, ..., 3.8719e-04, + 4.6921e-04, 4.5776e-04], + [-9.2773e-02, -9.1309e-02, -8.9844e-02, ..., -4.7874e-04, + -4.1199e-04, -4.4250e-04], + ..., + [-7.4707e-02, -7.2754e-02, -7.1777e-02, ..., 1.1444e-04, + 2.1648e-04, 1.1492e-04], + [-6.0791e-02, -5.9326e-02, -5.8594e-02, ..., 1.5736e-04, + 1.8477e-05, 2.9922e-05], + [ 7.2266e-02, 7.1289e-02, 7.0312e-02, ..., -9.0599e-05, + -1.1873e-04, -8.2016e-05]], dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_mlp.2.weight': tensor([[-4.9438e-03, -3.9978e-03, 2.5177e-03, ..., -7.4158e-03, + -1.1414e-02, 8.9111e-03], + [-3.3417e-03, -2.6550e-03, -1.6861e-03, ..., -2.7618e-03, + -2.5787e-03, 2.5787e-03], + [-6.4850e-05, -2.4414e-03, -1.7471e-03, ..., -3.5477e-04, + -1.5259e-04, 1.3161e-04], + ..., + [-4.6143e-02, -6.9275e-03, -5.6396e-02, ..., -3.4424e-02, + -1.2695e-02, 2.6489e-02], + [ 8.1177e-03, -2.7344e-02, 3.1738e-02, ..., 6.6528e-03, + 5.4016e-03, -6.5613e-03], + [ 9.4604e-03, 7.5989e-03, 4.9438e-03, ..., 8.5449e-03, + 4.1199e-03, -6.7444e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.after_proj.weight': tensor([[ 0.0049, 0.0079, -0.0066, ..., 0.0017, 0.0227, 0.0034], + [-0.0020, -0.0085, -0.0080, ..., 0.0134, -0.0012, -0.0102], + [-0.0118, 0.0030, -0.0039, ..., -0.0023, -0.0021, 0.0011], + ..., + [-0.0079, -0.0026, 0.0031, ..., -0.0070, 0.0021, -0.0082], + [-0.0040, 0.0076, 0.0044, ..., 0.0051, -0.0108, -0.0024], + [ 0.0119, 0.0110, 0.0015, ..., -0.0018, 0.0058, 0.0067]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.after_proj.bias': tensor([ 0.0050, -0.0118, -0.0096, ..., -0.0070, 0.0018, 0.0004], + dtype=torch.bfloat16)} +[06-26 08:28:41|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:41|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 08:28:41|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 1 checkpoint. +[06-26 08:28:41|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/depth(626e6618-bfcd-4d9a-a077-1409e2ce353f) +[06-26 08:28:41|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 2 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/dea7737ca29dd8d9086413c6dc5724b8250a0bb4/general/depth/626e6618-bfcd-4d9a-a077-1409e2ce353f_ema_bf16.pt +[06-26 08:28:42|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0144, -0.0021, 0.0064], + [ 0.0036, 0.0057, -0.0327, ..., -0.0082, -0.0024, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0137, -0.0029, 0.0283], + ..., + [ 0.0010, 0.0125, 0.0248, ..., 0.0437, -0.0074, -0.0192], + [ 0.0277, 0.0146, 0.0036, ..., -0.0031, -0.0092, 0.0146], + [ 0.0437, 0.0366, 0.0085, ..., 0.0013, -0.0128, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0028, -0.0310, 0.0300, ..., -0.0071, -0.0042, -0.0339], + [-0.0103, -0.0048, 0.0115, ..., 0.0049, 0.0092, 0.0520], + [-0.0064, 0.0132, 0.0398, ..., -0.0097, -0.0179, -0.0070], + ..., + [-0.0046, 0.0175, -0.0094, ..., 0.0225, 0.0047, -0.0020], + [-0.0209, -0.0081, 0.0087, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0354, -0.0082, ..., 0.0123, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0016, ..., 0.0376, 0.0532, 0.0011], + [ 0.0175, -0.0347, 0.0015, ..., -0.0064, 0.0047, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0156, 0.0262, -0.0114], + [-0.0015, 0.0244, 0.0356, ..., 0.0018, 0.0170, 0.0112], + [ 0.0486, 0.0175, 0.0042, ..., 0.0046, -0.0021, 0.0008]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0118, ..., -0.0075, -0.0337, 0.0079], + [ 0.0491, 0.0413, 0.0251, ..., -0.0041, 0.0393, 0.0035], + [-0.0162, -0.0005, 0.0134, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0014, 0.0261, 0.0042, ..., -0.0038, 0.0347, 0.0036], + [-0.0249, 0.0067, 0.0024, ..., 0.0096, 0.0388, -0.0205], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0123, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0048, -0.0270, -0.0166, ..., 0.0118, -0.0298, 0.0532], + [ 0.0034, 0.0036, -0.0126, ..., -0.0161, 0.0152, 0.0308], + [ 0.0145, 0.0086, 0.0137, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0067, -0.0630, 0.0027, ..., 0.0072, -0.0183, 0.0217], + [ 0.0039, 0.0032, 0.0258, ..., -0.0486, -0.0349, -0.0039], + [ 0.0182, -0.0081, 0.0197, ..., 0.0151, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0053, -0.0042, -0.0005, ..., -0.0007, -0.0017, -0.0121], + [-0.0152, 0.0251, 0.0432, ..., -0.0051, 0.0039, -0.0038], + [-0.0149, -0.0203, -0.0245, ..., -0.0118, -0.0325, 0.0029], + ..., + [ 0.0043, 0.0008, 0.0417, ..., 0.0280, 0.0022, -0.0024], + [-0.0039, 0.0037, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0067, -0.0034, -0.0085]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0155, -0.0063, -0.0481, ..., 0.0449, 0.0212, 0.0128], + [ 0.0156, -0.0227, 0.0220, ..., -0.0219, 0.0005, 0.0081], + [-0.0156, -0.0098, 0.0430, ..., -0.0101, 0.0327, -0.0110], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0136], + [ 0.0021, 0.0166, -0.0061, ..., 0.0013, 0.0079, -0.0052], + [-0.0110, -0.0041, 0.0117, ..., 0.0047, -0.0034, -0.0077]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0035, -0.0386, -0.0120, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0075, 0.0286, ..., -0.0339, -0.0160, 0.0170], + [-0.0127, -0.0053, -0.0128, ..., 0.0214, 0.0111, -0.0044], + ..., + [-0.0153, -0.0172, -0.0137, ..., 0.0098, 0.0098, 0.0184], + [-0.0154, -0.0356, 0.0201, ..., -0.0069, -0.0117, 0.0332], + [-0.0396, -0.0121, -0.0090, ..., -0.0284, -0.0141, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0280, 0.0312, 0.0133, ..., -0.0195, 0.0114, -0.0153], + [ 0.0192, 0.0092, -0.0073, ..., 0.0022, -0.0127, -0.0198], + [-0.0249, 0.0109, -0.0137, ..., 0.0090, 0.0172, 0.0067], + ..., + [ 0.0466, -0.0045, 0.0007, ..., 0.0010, 0.0239, 0.0042], + [ 0.0547, -0.0194, -0.0108, ..., -0.0069, 0.0311, -0.0044], + [ 0.0051, 0.0104, -0.0095, ..., 0.0203, 0.0170, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0044, -0.0046, 0.0283, ..., -0.0109, -0.0045, -0.0182], + [ 0.0052, 0.0126, -0.0315, ..., -0.0457, -0.0069, 0.0109], + [ 0.0006, 0.0261, 0.0245, ..., 0.0046, 0.0084, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0155, 0.0058, -0.0077], + [ 0.0059, -0.0371, 0.0356, ..., -0.0082, 0.0410, -0.0178], + [ 0.0178, -0.0155, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0087, 0.0208, 0.0272, ..., 0.0137, 0.0156, 0.0082], + [-0.0082, -0.0023, -0.0041, ..., 0.0144, 0.0007, 0.0057], + [ 0.0046, 0.0010, 0.0240, ..., -0.0119, -0.0066, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0046, ..., 0.0248, -0.0106, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., 0.0020, -0.0295, 0.0400], + [-0.0118, 0.0135, 0.0305, ..., -0.0100, 0.0366, 0.0227]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0054, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0149, -0.0054, 0.0288, ..., 0.0136, 0.0583, 0.0044], + [-0.0281, 0.0154, -0.0386, ..., -0.0444, -0.0093, -0.0009], + ..., + [ 0.0035, 0.0157, 0.0217, ..., 0.0047, 0.0115, 0.0151], + [ 0.0051, 0.0195, -0.0283, ..., -0.0048, -0.0039, -0.0342], + [-0.0167, -0.0195, 0.0007, ..., -0.0048, 0.0165, 0.0132]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0140, -0.0146, 0.0459, ..., 0.0043, -0.0320, 0.0275], + [ 0.0034, 0.0061, 0.0272, ..., -0.0219, -0.0183, -0.0265], + [ 0.0219, 0.0216, -0.0201, ..., -0.0022, 0.0052, 0.0041], + ..., + [-0.0258, 0.0028, 0.0215, ..., 0.0017, 0.0236, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0060], + [-0.0037, -0.0085, -0.0128, ..., -0.0334, 0.0062, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0146, -0.0454, ..., -0.0126, -0.0131, 0.0400], + [ 0.0295, -0.0187, -0.0106, ..., -0.0283, 0.0040, -0.0186], + [ 0.0013, 0.0547, -0.0042, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0280, 0.0045, -0.0269, ..., -0.0090, -0.0287, 0.0150], + [ 0.0093, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0098, 0.0106, ..., -0.0075, 0.0703, -0.0059]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0094, -0.0315, ..., 0.0058, 0.0254, 0.0359], + [ 0.0087, -0.0048, 0.0028, ..., -0.0125, -0.0233, -0.0042], + [ 0.0070, 0.0054, -0.0121, ..., -0.0123, 0.0018, 0.0089], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0040], + [ 0.0339, -0.0059, 0.0223, ..., 0.0410, 0.0028, 0.0138], + [-0.0055, -0.0439, -0.0486, ..., 0.0140, -0.0130, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0002, 0.0139, -0.0042, ..., -0.0079, -0.0045, -0.0229], + [-0.0289, 0.0231, 0.0084, ..., -0.0140, 0.0056, 0.0222], + [ 0.0085, 0.0295, -0.0193, ..., -0.0164, 0.0310, -0.0262], + ..., + [ 0.0233, -0.0053, -0.0029, ..., 0.0051, 0.0278, 0.0031], + [-0.0104, -0.0018, 0.0140, ..., -0.0175, 0.0444, 0.0425], + [ 0.0054, -0.0084, 0.0096, ..., 0.0061, -0.0344, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[ 0.0027, 0.0126, 0.0188, ..., -0.0092, -0.0199, 0.0147], + [-0.0581, 0.0294, -0.0043, ..., -0.0240, 0.0058, 0.0302], + [ 0.0087, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [ 0.0002, -0.0292, 0.0537, ..., 0.0033, -0.0078, -0.0106], + [-0.0036, -0.0077, 0.0209, ..., 0.0043, 0.0093, 0.0187], + [-0.0087, -0.0349, -0.0173, ..., -0.0400, 0.0057, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0182, -0.0265, -0.0105, ..., 0.0099, -0.0132, -0.0405], + [ 0.0051, -0.0087, 0.0266, ..., -0.0400, 0.0060, 0.0391], + [-0.0016, 0.0359, 0.0284, ..., 0.0292, -0.0233, 0.0212], + ..., + [ 0.0026, 0.0121, 0.0132, ..., -0.0141, -0.0048, 0.0095], + [ 0.0299, -0.0167, -0.0042, ..., -0.0013, -0.0002, -0.0206], + [ 0.0072, -0.0045, -0.0042, ..., 0.0037, -0.0022, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, -0.0015, 0.0118, ..., -0.0029, -0.0208, -0.0074], + [-0.0096, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0232], + [ 0.0134, -0.0278, 0.0118, ..., 0.0396, -0.0079, -0.0069], + ..., + [-0.0001, -0.0018, -0.0311, ..., 0.0344, 0.0051, -0.0123], + [-0.0236, -0.0102, -0.0072, ..., 0.0332, -0.0114, -0.0179], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0137, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[-0.0014, 0.0166, 0.0036, ..., 0.0273, -0.0352, 0.0156], + [-0.0510, -0.0062, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0077, ..., -0.0388, -0.0019, 0.0398], + ..., + [-0.0160, -0.0195, -0.0027, ..., 0.0037, -0.0056, 0.0029], + [-0.0187, 0.0038, 0.0098, ..., 0.0287, -0.0125, -0.0095], + [-0.0181, -0.0032, 0.0247, ..., -0.0143, 0.0052, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0096, 0.0141, ..., -0.0112, 0.0026, -0.0253], + [-0.0167, 0.0330, -0.0171, ..., -0.0037, -0.0403, -0.0110], + [ 0.0084, 0.0347, 0.0275, ..., 0.0148, 0.0064, 0.0079], + ..., + [ 0.0272, -0.0168, -0.0106, ..., -0.0098, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0059, ..., -0.0032, -0.0018, -0.0243], + [-0.0075, 0.0581, 0.0057, ..., 0.0145, 0.0239, 0.0116]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0118, -0.0410, 0.0126, ..., -0.0254, -0.0052, -0.0173], + [ 0.0476, 0.0315, 0.0148, ..., -0.0150, -0.0066, -0.0012], + [-0.0089, -0.0269, 0.0083, ..., 0.0332, 0.0144, -0.0063], + ..., + [-0.0046, -0.0043, 0.0160, ..., -0.0177, 0.0168, -0.0126], + [ 0.0051, -0.0187, -0.0049, ..., 0.0248, -0.0148, -0.0068], + [-0.0142, -0.0024, 0.0148, ..., -0.0043, 0.0037, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0054, 0.0094, ..., 0.0255, -0.0253, -0.0135], + [-0.0018, -0.0143, -0.0160, ..., -0.0113, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0115, 0.0046, 0.0322, ..., -0.0183, -0.0010, 0.0016], + [ 0.0034, -0.0050, 0.0162, ..., 0.0043, 0.0255, 0.0063], + [ 0.0270, 0.0087, 0.0159, ..., 0.0189, 0.0148, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0066, ..., -0.0532, 0.0104, 0.0036], + [-0.0240, -0.0215, -0.0103, ..., -0.0059, 0.0098, -0.0168], + [ 0.0074, -0.0013, -0.0059, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0057, 0.0172, -0.0106, ..., -0.0300, -0.0045, 0.0435], + [ 0.0051, -0.0261, 0.0247, ..., 0.0192, -0.0308, 0.0061], + [-0.0167, 0.0123, 0.0106, ..., -0.0247, 0.0123, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0078, -0.0078, 0.0317, ..., 0.0200, -0.0104, -0.0376], + [ 0.0356, 0.0378, 0.0135, ..., 0.0325, 0.0391, 0.0173], + [-0.0244, -0.0264, 0.0371, ..., 0.0028, 0.0052, -0.0322], + ..., + [ 0.0058, 0.0107, -0.0040, ..., 0.0366, -0.0206, 0.0374], + [-0.0046, -0.0070, 0.0034, ..., -0.0093, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0026, ..., -0.0166, -0.0056, 0.0245]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0053, -0.0093, -0.0344, ..., -0.0076, 0.0347, -0.0176], + [ 0.0267, -0.0366, -0.0088, ..., 0.0217, 0.0349, -0.0043], + [-0.0097, -0.0286, 0.0074, ..., 0.0052, -0.0282, 0.0013], + ..., + [-0.0129, 0.0037, 0.0058, ..., -0.0086, 0.0165, 0.0076], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0129], + [-0.0140, 0.0043, -0.0170, ..., -0.0127, 0.0060, 0.0070]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0025, -0.0047, ..., -0.0483, 0.0374, 0.0093], + [-0.0576, 0.0116, 0.0374, ..., -0.0153, -0.0188, -0.0136], + ..., + [ 0.0260, -0.0007, -0.0077, ..., -0.0003, 0.0053, -0.0165], + [-0.0222, -0.0155, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0131, -0.0148, -0.0038, ..., 0.0018, -0.0234, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0101], + [ 0.0249, 0.0205, 0.0075, ..., -0.0061, -0.0089, -0.0105], + [-0.0220, 0.0044, 0.0047, ..., 0.0309, 0.0381, 0.0037], + ..., + [ 0.0200, -0.0033, 0.0171, ..., 0.0122, 0.0320, -0.0157], + [ 0.0349, 0.0120, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0075, 0.0256, 0.0050, ..., 0.0205, 0.0035, -0.0105]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0043, 0.0199, ..., -0.0050, 0.0276, -0.0308], + [ 0.0178, -0.0063, 0.0099, ..., 0.0108, -0.0398, 0.0300], + [ 0.0430, -0.0124, 0.0081, ..., -0.0009, -0.0066, -0.0055], + ..., + [ 0.0192, -0.0106, -0.0108, ..., -0.0025, 0.0330, 0.0142], + [-0.0124, -0.0542, -0.0205, ..., 0.0062, -0.0159, 0.0320], + [ 0.0303, 0.0054, -0.0425, ..., 0.0009, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0068, -0.0315, ..., 0.0525, 0.0034, -0.0349], + [-0.0044, -0.0070, -0.0156, ..., 0.0053, -0.0240, 0.0305], + [-0.0256, -0.0045, 0.0062, ..., -0.0459, -0.0148, -0.0165], + ..., + [ 0.0070, 0.0231, 0.0209, ..., 0.0077, -0.0038, 0.0461], + [-0.0067, -0.0085, 0.0004, ..., -0.0217, 0.0347, -0.0018], + [ 0.0057, 0.0234, 0.0056, ..., 0.0261, -0.0038, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0071, 0.0023], + [-0.0183, -0.0211, 0.0045, ..., -0.0145, 0.0143, -0.0166], + [-0.0075, -0.0160, 0.0053, ..., -0.0203, 0.0267, 0.0217], + ..., + [ 0.0308, 0.0096, 0.0075, ..., 0.0171, 0.0053, -0.0070], + [-0.0145, 0.0260, -0.0085, ..., 0.0077, 0.0102, -0.0166], + [ 0.0094, -0.0142, -0.0037, ..., 0.0051, 0.0069, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0070, 0.0130, -0.0076, ..., -0.0278, 0.0240, -0.0126], + [ 0.0035, 0.0187, -0.0160, ..., -0.0317, -0.0032, 0.0173], + [ 0.0062, 0.0310, -0.0092, ..., -0.0226, 0.0030, 0.0334], + ..., + [-0.0079, 0.0186, 0.0038, ..., 0.0228, 0.0069, 0.0065], + [ 0.0271, 0.0117, 0.0354, ..., -0.0076, 0.0198, 0.0145], + [ 0.0282, 0.0171, 0.0317, ..., -0.0111, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0114, -0.0265, -0.0110, ..., 0.0049, 0.0037, 0.0136], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0085], + [-0.0067, -0.0190, -0.0045, ..., 0.0166, -0.0135, 0.0138], + ..., + [-0.0200, -0.0012, 0.0061, ..., -0.0339, 0.0361, -0.0025], + [ 0.0037, -0.0250, 0.0153, ..., 0.0006, -0.0162, 0.0220], + [-0.0078, 0.0028, 0.0205, ..., -0.0064, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0053, 0.0240, -0.0217, ..., 0.0172, 0.0048, 0.0114], + [-0.0386, 0.0103, 0.0075, ..., -0.0042, 0.0322, 0.0086], + [-0.0427, 0.0040, 0.0035, ..., 0.0073, 0.0110, -0.0137], + ..., + [-0.0087, -0.0041, -0.0264, ..., -0.0586, -0.0630, -0.0132], + [ 0.0045, 0.0287, 0.0420, ..., -0.0311, -0.0090, -0.0289], + [-0.0107, -0.0610, 0.0162, ..., -0.0291, -0.0166, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0121, -0.0251, 0.0211, ..., 0.0199, -0.0084, -0.0112], + [ 0.0112, -0.0099, -0.0327, ..., 0.0447, 0.0076, 0.0074], + [ 0.0035, -0.0175, 0.0066, ..., -0.0062, -0.0049, -0.0065], + ..., + [ 0.0295, 0.0045, -0.0104, ..., 0.0093, -0.0160, -0.0288], + [ 0.0083, 0.0253, -0.0177, ..., 0.0094, 0.0074, 0.0080], + [-0.0129, 0.0027, 0.0137, ..., -0.0128, 0.0277, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0131, 0.0047, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0068, 0.0234, 0.0065, ..., 0.0464, 0.0064, 0.0309], + ..., + [-0.0117, -0.0175, 0.0327, ..., 0.0132, 0.0044, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, 0.0059, -0.0444, ..., 0.0060, -0.0283, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0082, -0.0099, 0.0146, ..., -0.0742, -0.0064, 0.0161], + [-0.0076, -0.0131, -0.0075, ..., 0.0064, 0.0026, 0.0238], + [ 0.0189, 0.0221, -0.0088, ..., 0.0050, 0.0408, 0.0052], + ..., + [ 0.0059, 0.0247, 0.0396, ..., 0.0064, -0.0304, -0.0183], + [ 0.0299, -0.0114, -0.0417, ..., -0.0048, -0.0376, -0.0044], + [-0.0210, 0.0168, -0.0026, ..., 0.0049, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0069, 0.0242, 0.0062, ..., -0.0245, 0.0078, 0.0002], + [ 0.0190, -0.0125, 0.0332, ..., -0.0146, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0068, ..., -0.0059, -0.0092, 0.0182], + ..., + [-0.0430, -0.0403, -0.0081, ..., 0.0135, -0.0140, 0.0046], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0094, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0042, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0092, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0067, ..., 0.0101, -0.0223, 0.0062], + ..., + [ 0.0192, 0.0292, 0.0059, ..., 0.0176, -0.0197, 0.0046], + [ 0.0265, 0.0063, 0.0054, ..., 0.0286, -0.0021, -0.0267], + [-0.0184, -0.0088, -0.0048, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0084, -0.0159, 0.0082, ..., 0.0198, 0.0056, 0.0277], + [ 0.0172, 0.0041, -0.0057, ..., 0.0048, -0.0049, 0.0057], + [ 0.0036, 0.0184, 0.0113, ..., 0.0099, -0.0056, -0.0074], + ..., + [ 0.0048, -0.0111, 0.0035, ..., -0.0096, -0.0053, 0.0122], + [-0.0427, -0.0391, -0.0192, ..., 0.0091, -0.0238, -0.0078], + [ 0.0055, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0084, -0.0161, 0.0123], + [-0.0081, 0.0060, 0.0312, ..., 0.0053, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0071, -0.0014, 0.0278], + ..., + [ 0.0170, -0.0085, 0.0471, ..., 0.0067, 0.0078, -0.0073], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0076, -0.0203], + [ 0.0053, 0.0048, -0.0081, ..., -0.0095, -0.0086, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[-0.0058, -0.0156, 0.0284, ..., -0.0104, 0.0135, -0.0044], + [ 0.0152, 0.0332, 0.0337, ..., 0.0223, -0.0165, -0.0044], + [ 0.0055, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [-0.0056, -0.0019, 0.0280, ..., -0.0076, -0.0078, 0.0140], + [-0.0083, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0091, ..., 0.0067, 0.0167, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0042, ..., -0.0238, -0.0103, 0.0119], + [-0.0250, -0.0162, 0.0034, ..., -0.0081, 0.0002, -0.0369], + [ 0.0051, 0.0141, 0.0079, ..., -0.0452, -0.0038, 0.0120], + ..., + [ 0.0148, -0.0141, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0101, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0142, 0.0041, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[ 0.0028, -0.0327, -0.0190, ..., 0.0064, -0.0052, -0.0209], + [ 0.0713, 0.0152, -0.0052, ..., -0.0049, 0.0186, 0.0027], + [-0.0227, -0.0092, 0.0188, ..., -0.0122, 0.0193, 0.0053], + ..., + [ 0.0116, 0.0464, 0.0085, ..., 0.0248, 0.0168, 0.0047], + [ 0.0381, -0.0038, 0.0071, ..., -0.0031, -0.0107, -0.0054], + [ 0.0292, 0.0057, 0.0209, ..., 0.0056, 0.0234, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0056, 0.0212, ..., 0.0305, 0.0137, 0.0105], + [-0.0078, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0045], + ..., + [ 0.0084, -0.0275, 0.0225, ..., 0.0043, -0.0189, -0.0107], + [-0.0128, 0.0053, -0.0058, ..., -0.0079, 0.0064, 0.0211], + [ 0.0015, -0.0073, -0.0063, ..., -0.0109, 0.0264, 0.0059]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[-0.0049, -0.0075, -0.0108, ..., 0.0186, -0.0216, -0.0047], + [ 0.0048, 0.0123, -0.0231, ..., 0.0123, -0.0112, -0.0183], + [ 0.0107, 0.0103, -0.0057, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0125, 0.0154, ..., 0.0540, 0.0415, 0.0097], + [ 0.0025, 0.0087, -0.0137, ..., -0.0030, -0.0189, -0.0237], + [ 0.0059, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0076, 0.0359, ..., -0.0084, -0.0120, -0.0049], + [ 0.0190, -0.0059, 0.0298, ..., 0.0126, -0.0050, -0.0200], + [ 0.0112, -0.0110, 0.0177, ..., -0.0248, 0.0229, 0.0034], + ..., + [-0.0081, 0.0044, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0044, -0.0291, ..., 0.0206, 0.0304, -0.0085], + [-0.0080, 0.0245, -0.0068, ..., -0.0413, 0.0216, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0078, -0.0157, 0.0043, ..., -0.0052, 0.0228, -0.0339], + [-0.0713, -0.0153, -0.0168, ..., -0.0217, -0.0183, -0.0073], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, -0.0063], + ..., + [-0.0027, -0.0069, 0.0258, ..., -0.0310, 0.0091, -0.0361], + [ 0.0063, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [-0.0071, 0.0157, -0.0069, ..., 0.0075, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0077, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0106, ..., 0.0104, -0.0210, 0.0148], + [-0.0286, -0.0267, -0.0037, ..., 0.0129, -0.0059, 0.0087], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [-0.0062, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0129, 0.0206, ..., 0.0069, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0152, -0.0092, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0045, 0.0076, ..., -0.0046, 0.0110, 0.0598], + [-0.0062, 0.0267, -0.0053, ..., 0.0284, -0.0045, -0.0364], + ..., + [ 0.0045, 0.0078, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0017, -0.0087, ..., 0.0211, 0.0137, -0.0075], + [ 0.0101, 0.0236, -0.0206, ..., -0.0136, -0.0063, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0067, 0.0294, ..., 0.0040, 0.0270, 0.0065], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0066, -0.0168], + [ 0.0201, -0.0121, 0.0126, ..., -0.0381, -0.0247, -0.0059], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0078], + [-0.0194, 0.0094, -0.0092, ..., -0.0231, -0.0069, -0.0466], + [ 0.0327, 0.0284, -0.0044, ..., 0.0145, 0.0142, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, -0.0021, -0.0167, ..., -0.0037, 0.0170, 0.0107], + [ 0.0019, -0.0031, 0.0295, ..., -0.0172, 0.0040, 0.0135], + [ 0.0193, 0.0236, -0.0125, ..., 0.0038, -0.0369, 0.0239], + ..., + [ 0.0151, -0.0413, -0.0124, ..., -0.0354, 0.0065, -0.0320], + [-0.0266, -0.0063, -0.0327, ..., -0.0050, 0.0058, 0.0073], + [ 0.0298, 0.0356, -0.0192, ..., 0.0042, 0.0008, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0045, 0.0058, 0.0317, ..., 0.0062, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0071, -0.0096], + [-0.0166, 0.0064, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0041, 0.0260, ..., 0.0159, -0.0035, -0.0474], + [-0.0064, 0.0079, 0.0076, ..., 0.0140, 0.0557, 0.0052], + [-0.0203, 0.0273, -0.0040, ..., -0.0330, -0.0038, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-3.3203e-02, -1.2390e-02, -3.8757e-03, ..., 6.3782e-03, + -9.3384e-03, -2.7710e-02], + [ 4.9744e-03, -3.4637e-03, -5.5237e-03, ..., 6.5002e-03, + 1.4038e-02, -1.6113e-02], + [-4.6997e-03, 1.6846e-02, -7.6599e-03, ..., -9.1553e-03, + 1.2390e-02, 2.0447e-03], + ..., + [ 3.9795e-02, 4.5166e-02, 2.5146e-02, ..., -5.8289e-03, + -3.2959e-02, 4.1199e-03], + [-6.2012e-02, 3.5248e-03, 3.7109e-02, ..., 4.1016e-02, + 9.2506e-05, -1.7334e-02], + [ 1.2634e-02, 5.7678e-03, -2.6978e-02, ..., 4.9438e-03, + 3.8818e-02, -1.0376e-02]], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0013, 0.0245], + [-0.0449, 0.0019, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, 0.0062, ..., 0.0041, 0.0288, 0.0124], + [ 0.0330, 0.0038, 0.0060, ..., 0.0299, 0.0142, -0.0483], + [-0.0166, 0.0077, 0.0051, ..., -0.0062, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0045, -0.0025, 0.0527], + [-0.0002, 0.0403, 0.0112, ..., 0.0417, 0.0105, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0096, -0.0096, 0.0327], + [ 0.0142, 0.0111, -0.0069, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0124, 0.0281, ..., -0.0242, -0.0127, -0.0047]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0042], + [ 0.0063, -0.0045, -0.0286, ..., 0.0056, 0.0172, 0.0106], + [ 0.0393, 0.0400, 0.0039, ..., 0.0025, 0.0082, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0107, 0.0097, 0.0150], + [-0.0057, 0.0039, 0.0077, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0078, ..., 0.0092, -0.0114, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0071, 0.0041, ..., -0.0220, -0.0145, -0.0013], + [-0.0166, 0.0189, -0.0047, ..., 0.0096, 0.0020, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0054, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0057, ..., 0.0047, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0082, 0.0071, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0122, 0.0289, 0.0125, ..., 0.0135, 0.0070, -0.0104], + [ 0.0188, -0.0197, 0.0010, ..., 0.0284, -0.0342, -0.0057], + [-0.0525, -0.0215, 0.0154, ..., -0.0270, 0.0081, 0.0254], + ..., + [-0.0069, 0.0087, 0.0281, ..., 0.0129, -0.0064, -0.0048], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0045, 0.0038], + [-0.0273, -0.0747, 0.0208, ..., -0.0044, -0.0044, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0048, 0.0150, 0.0334, ..., 0.0205, -0.0273, -0.0037], + [-0.0165, 0.0111, -0.0135, ..., 0.0106, -0.0071, 0.0219], + [ 0.0063, 0.0227, -0.0054, ..., 0.0128, -0.0077, 0.0150], + ..., + [-0.0347, -0.0148, 0.0053, ..., -0.0046, -0.0042, -0.0094], + [-0.0077, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0069, -0.0500, -0.0232, ..., 0.0320, 0.0057, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0056, -0.0364], + [ 0.0220, -0.0026, -0.0112, ..., -0.0052, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, 0.0039, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0064, 0.0192, -0.0393], + [-0.0102, -0.0244, 0.0275, ..., 0.0029, 0.0042, 0.0081], + [-0.0170, -0.0427, 0.0140, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0040, -0.0378, -0.0359, ..., 0.0552, -0.0062, -0.0160], + [ 0.0039, 0.0143, 0.0134, ..., 0.0049, -0.0153, 0.0354], + [ 0.0095, -0.0038, 0.0065, ..., -0.0046, -0.0249, -0.0136], + ..., + [ 0.0081, -0.0124, 0.0050, ..., -0.0095, -0.0085, 0.0047], + [-0.0051, 0.0439, 0.0166, ..., 0.0248, 0.0131, -0.0187], + [ 0.0175, 0.0059, 0.0041, ..., 0.0410, -0.0096, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0104, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0141, -0.0050, ..., 0.0052, -0.0244, 0.0356], + [ 0.0063, 0.0165, -0.0200, ..., -0.0019, 0.0237, -0.0078], + ..., + [-0.0197, -0.0049, 0.0386, ..., 0.0181, 0.0112, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, 0.0041, -0.0374], + [-0.0086, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0066, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0135, 0.0114, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, 0.0045, -0.0041, ..., 0.0050, -0.0055, 0.0337], + ..., + [ 0.0422, 0.0033, 0.0352, ..., -0.0304, -0.0049, -0.0204], + [ 0.0302, 0.0156, 0.0099, ..., -0.0078, -0.0254, -0.0245], + [-0.0040, -0.0060, -0.0039, ..., 0.0037, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0052, 0.0266, ..., 0.0154, 0.0187, 0.0047], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0051, ..., -0.0172, -0.0091, 0.0231], + ..., + [-0.0295, -0.0137, 0.0038, ..., 0.0043, -0.0044, 0.0099], + [-0.0049, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0078], + [-0.0040, -0.0048, -0.0259, ..., -0.0039, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0079, 0.0078, -0.0140, ..., 0.0242, -0.0070, -0.0386], + [-0.0140, 0.0079, -0.0097, ..., -0.0369, -0.0156, 0.0157], + [-0.0267, 0.0078, -0.0057, ..., -0.0126, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0007, 0.0144, ..., 0.0115, 0.0042, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., -0.0046, 0.0042, -0.0137], + [-0.0219, -0.0396, 0.0041, ..., -0.0058, -0.0018, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0034, 0.0168, 0.0194, ..., 0.0349, -0.0420, 0.0017], + [ 0.0042, 0.0273, 0.0145, ..., -0.0184, -0.0039, 0.0154], + [ 0.0145, -0.0036, -0.0072, ..., 0.0245, 0.0359, 0.0059], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0046, 0.0033, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0070, -0.0312, -0.0457], + [ 0.0050, 0.0054, -0.0053, ..., 0.0049, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0015, 0.0123, -0.0029, ..., 0.0124, -0.0038, 0.0066], + [-0.0277, -0.0036, -0.0181, ..., -0.0027, -0.0034, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0042, 0.0513, -0.0110], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0111], + [-0.0282, -0.0053, 0.0121, ..., -0.0051, -0.0039, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0039, -0.0128, ..., 0.0140, 0.0308, 0.0058], + [ 0.0119, -0.0332, -0.0320, ..., -0.0078, -0.0140, -0.0029], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, -0.0028, -0.0311, ..., 0.0120, 0.0061, 0.0081], + [ 0.0048, 0.0198, 0.0142, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, -0.0046, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0091, -0.0036, 0.0056, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0018, ..., 0.0302, -0.0024, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0039, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0025, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0049, -0.0222, 0.0262], + ..., + [-0.0147, 0.0058, 0.0260, ..., 0.0051, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0057, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0076, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0113], + [-0.0278, 0.0044, -0.0374, ..., 0.0276, -0.0038, 0.0151], + [ 0.0089, 0.0237, -0.0067, ..., 0.0527, -0.0070, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0068, -0.0095], + [-0.0075, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0044, -0.0128, ..., 0.0049, -0.0371, 0.0017], + [ 0.0075, -0.0126, 0.0049, ..., -0.0204, 0.0041, 0.0056], + [-0.0110, -0.0195, 0.0056, ..., -0.0542, -0.0092, -0.0053], + ..., + [-0.0205, -0.0085, -0.0041, ..., -0.0199, 0.0041, 0.0061], + [-0.0220, -0.0302, -0.0099, ..., 0.0104, 0.0369, 0.0043], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[-0.0023, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0047, 0.0049, 0.0238, ..., -0.0145, -0.0320, 0.0037], + [-0.0193, 0.0364, -0.0048, ..., -0.0064, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0048, -0.0232, -0.0464], + [-0.0109, -0.0298, -0.0195, ..., 0.0114, 0.0312, 0.0156], + [-0.0074, -0.0117, -0.0195, ..., 0.0043, 0.0106, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0048, -0.0239, -0.0123, ..., -0.0045, 0.0243, -0.0396], + [-0.0101, 0.0155, -0.0056, ..., -0.0277, -0.0038, 0.0037], + ..., + [ 0.0117, -0.0057, 0.0515, ..., 0.0327, -0.0120, -0.0066], + [ 0.0041, 0.0170, 0.0074, ..., -0.0048, 0.0280, -0.0210], + [ 0.0036, 0.0068, 0.0057, ..., 0.0153, -0.0077, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0061, 0.0081, -0.0077, ..., -0.0588, 0.0187, -0.0064], + [-0.0102, 0.0199, -0.0077, ..., 0.0080, -0.0332, 0.0040], + ..., + [ 0.0078, -0.0464, 0.0277, ..., -0.0145, 0.0198, -0.0078], + [ 0.0162, 0.0038, 0.0090, ..., -0.0109, 0.0081, 0.0297], + [-0.0269, 0.0135, 0.0164, ..., 0.0079, 0.0176, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0042, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, -0.0026, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, 0.0048, -0.0201, ..., 0.0330, 0.0309, -0.0051], + ..., + [ 0.0039, 0.0123, 0.0153, ..., 0.0039, -0.0039, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0040, 0.0175], + [-0.0244, -0.0059, -0.0300, ..., 0.0049, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0045, -0.0041, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0038, ..., 0.0310, 0.0088, -0.0039], + [ 0.0025, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0110, -0.0237, ..., -0.0104, 0.0157, 0.0078], + [-0.0070, 0.0205, 0.0143, ..., 0.0262, -0.0195, 0.0095], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0078]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0024, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0022, 0.0034, -0.0093], + [ 0.0017, -0.0010, 0.0398, ..., -0.0304, -0.0028, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0039], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0051, 0.0291], + [ 0.0132, 0.0038, -0.0339, ..., 0.0077, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, -0.0027], + [-0.0166, -0.0036, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0061, -0.0032, ..., 0.0240, 0.0036, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0019], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0007, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0067, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0040], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, 0.0040, -0.0039], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, -0.0039, ..., -0.0164, -0.0039, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0026, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0025, 0.0226], + ..., + [ 0.0053, -0.0231, -0.0255, ..., -0.0052, 0.0295, -0.0058], + [-0.0255, 0.0513, 0.0039, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0040, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0056, 0.0078, -0.0364, ..., -0.0175, 0.0306, -0.0075], + [-0.0051, -0.0195, -0.0016, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0070, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0078, -0.0193, -0.0140], + [-0.0166, 0.0059, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0138, 0.0082, 0.0146, ..., -0.0081, 0.0211, -0.0270], + [ 0.0063, -0.0077, -0.0234, ..., 0.0089, -0.0055, -0.0165], + [ 0.0225, -0.0251, 0.0047, ..., 0.0072, 0.0236, -0.0036], + ..., + [ 0.0186, 0.0148, -0.0053, ..., 0.0272, -0.0062, -0.0086], + [ 0.0064, 0.0029, 0.0078, ..., -0.0260, 0.0267, 0.0147], + [ 0.0101, -0.0044, 0.0212, ..., 0.0195, -0.0157, 0.0114]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0084, -0.0270, 0.0251, ..., 0.0098, 0.0055, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.2.proj.1.weight': tensor([[ 0.0036, 0.0231, 0.0918, ..., -0.0571, 0.0250, -0.0284], + [-0.0237, 0.0006, -0.1128, ..., 0.0193, -0.0238, 0.0518], + [-0.0087, 0.1279, -0.0123, ..., -0.0220, 0.0184, -0.0349], + ..., + [ 0.0146, 0.0055, -0.0076, ..., 0.0234, -0.0280, 0.0082], + [-0.0237, -0.0608, 0.0603, ..., 0.0332, 0.0708, 0.0505], + [ 0.0146, -0.0229, -0.0050, ..., 0.0249, 0.0420, -0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_proj.weight': tensor([[-0.0093, 0.0068, 0.0118, ..., -0.0078, -0.0047, 0.0182], + [ 0.0018, 0.0015, -0.0032, ..., -0.0032, -0.0005, -0.0092], + [-0.0064, 0.0067, -0.0002, ..., 0.0002, -0.0104, -0.0227], + ..., + [ 0.0010, -0.0172, -0.0045, ..., 0.0082, 0.0030, -0.0064], + [-0.0014, 0.0045, -0.0045, ..., -0.0121, -0.0164, -0.0048], + [ 0.0080, -0.0050, -0.0016, ..., 0.0063, 0.0061, 0.0010]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_norm.weight': tensor([ 1.4062e-01, -4.2480e-02, 8.6426e-02, 2.3145e-01, -3.4485e-03, + -8.3160e-04, -1.3933e-06, -1.8794e-21, 6.6174e-22, -2.0117e-20, + -9.3408e-33, 3.3352e-21, -6.1512e-05, -1.1942e-23, -2.9565e-25, + 2.0954e-30, 1.8597e-05, 2.6484e-09, -3.7517e-12, -1.8406e-04, + 2.4738e-10, -4.5528e-21, 4.1504e-02, -3.2616e-04, -1.0742e-01, + 5.8594e-01, 5.4932e-02, -7.3853e-03, 4.9805e-02, 9.8633e-02, + -6.3019e-19, -1.4095e-21, 1.0757e-03, -9.5367e-05, 1.2346e-13, + -2.4033e-04, 3.4106e-12, -7.1086e-25, 1.1529e-23, -2.0292e-24, + -2.3588e-25, -6.0988e-26, 1.1228e-25, -6.5308e-03, 4.2969e-02, + -1.4771e-02, 2.9785e-02, 1.2354e-01, 3.4180e-01, 7.8125e-03, + -4.6082e-03, 5.3406e-03, 4.4556e-03, 1.2005e-10, 2.8922e-10, + 1.4095e-17, -4.8307e-22, -1.0503e-18, 3.0494e-26, -2.6496e-24, + 5.1670e-29, -8.8938e-20, 2.1487e-25, 2.4536e-14, 2.7148e-01, + 2.2559e-01, 3.7109e-01, -4.9316e-02, 2.6855e-03, 7.1335e-04, + 1.0282e-06, 4.3675e-22, 1.2811e-20, 5.7175e-20, -1.3482e-31, + -7.3123e-22, 3.6163e-03, -7.3412e-24, -2.1306e-26, 6.2320e-29, + 1.9226e-03, -9.7316e-11, 4.0927e-11, -3.8452e-03, 3.3253e-12, + -6.6746e-19, -3.9978e-03, -1.2360e-03, 5.7812e-01, 1.1719e-01, + 5.0391e-01, 9.2773e-02, -5.8594e-03, 9.0408e-04, 1.6263e-18, + -5.2675e-21, 3.0060e-03, 4.6730e-05, 4.2188e-14, 3.9978e-03, + -5.7116e-10, -1.4217e-24, -8.4011e-25, -2.8290e-22, -4.2914e-28, + -2.5022e-23, 8.5591e-29, 1.4404e-02, 5.4626e-03, -6.0425e-03, + 4.1211e-01, 2.7734e-01, 2.0410e-01, -1.1047e-02, -5.2185e-03, + 7.8125e-03, 4.6692e-03, -4.8885e-11, 1.4461e-10, -9.1480e-19, + 4.1558e-21, 1.2197e-18, 3.3120e-26, -3.9911e-23, -1.2572e-29, + -3.2591e-22, -1.7707e-24, 8.5038e-11], dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.self_attn.k_proj.weight': tensor([[-0.0053, 0.0036, -0.0114, ..., -0.0015, -0.0072, 0.0092], + [-0.0233, -0.0015, 0.0042, ..., -0.0058, -0.0069, -0.0112], + [-0.0036, -0.0111, -0.0016, ..., 0.0052, 0.0021, 0.0089], + ..., + [ 0.0049, -0.0039, -0.0018, ..., -0.0025, 0.0008, 0.0016], + [ 0.0049, 0.0009, 0.0007, ..., -0.0044, 0.0020, 0.0024], + [ 0.0040, 0.0023, 0.0050, ..., -0.0034, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.k_norm.weight': tensor([ 2.4121e-01, 1.2598e-01, 6.2012e-02, 2.0898e-01, 3.3417e-03, + -1.1292e-03, 1.4082e-06, -1.9852e-21, 2.2896e-21, 2.8164e-20, + 1.9548e-32, 3.2955e-21, -9.9659e-05, -1.1581e-23, -2.9404e-25, + 2.7364e-30, -7.3016e-07, 2.6921e-09, 2.0464e-12, 1.9360e-04, + -2.2555e-10, 2.7926e-21, 3.7354e-02, -9.9182e-04, 7.6172e-02, + 5.8594e-01, 2.2852e-01, 8.9844e-02, 6.2256e-03, 6.8970e-03, + 2.5072e-19, 9.3306e-22, -3.7003e-04, -9.8228e-05, 1.4122e-13, + -1.0824e-04, -3.0695e-11, -1.3248e-25, -1.1684e-23, 1.0792e-24, + -2.2941e-25, -1.1551e-25, -1.1228e-25, 1.0681e-02, 4.4678e-02, + -1.1536e-02, 2.2852e-01, 2.0996e-01, 3.7109e-01, 1.2085e-02, + -4.2725e-03, -7.8125e-03, 7.6904e-03, -1.2824e-10, 3.1650e-10, + -7.5352e-18, -1.7999e-21, 1.3281e-18, 4.2005e-26, -2.7271e-24, + -5.4037e-29, -8.6821e-20, -2.1487e-25, 2.6867e-14, -1.0547e-01, + 2.2168e-01, 3.7305e-01, -1.5430e-01, -2.1667e-03, -9.6321e-05, + -1.0431e-06, -6.3693e-23, 1.2705e-20, 5.5904e-20, 1.3173e-31, + 1.0059e-21, -3.5858e-03, 6.2814e-24, -2.0700e-26, 6.6659e-29, + 1.9302e-03, 2.0191e-10, 4.2292e-11, -3.8452e-03, -2.7427e-12, + -6.5391e-19, -7.8735e-03, -2.8229e-04, 5.7812e-01, -1.3477e-01, + 4.9609e-01, 7.0801e-03, 4.9072e-02, 9.4727e-02, -1.8431e-18, + -5.2675e-21, -3.1891e-03, 2.1338e-05, -1.9207e-14, -3.9978e-03, + 5.8208e-10, -1.4282e-24, 3.2829e-24, -2.7132e-22, 1.6787e-27, + -1.5303e-23, 2.9819e-28, 9.4604e-03, 2.0294e-03, 1.0559e-02, + 4.0430e-01, 2.3340e-01, 1.8555e-01, 8.2397e-03, 5.9814e-03, + 4.6387e-03, -1.2054e-03, 1.6257e-11, 2.6489e-11, -8.4026e-18, + 3.0308e-21, -2.2192e-19, -2.1104e-26, 3.8877e-23, -1.0748e-29, + 6.0880e-21, -1.8482e-24, -8.4583e-11], dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.self_attn.v_proj.weight': tensor([[-0.0007, 0.0037, -0.0014, ..., 0.0045, 0.0013, -0.0028], + [ 0.0083, 0.0014, -0.0035, ..., -0.0039, 0.0026, 0.0009], + [ 0.0136, 0.0058, 0.0026, ..., -0.0038, -0.0015, 0.0028], + ..., + [-0.0036, -0.0049, 0.0043, ..., 0.0056, -0.0034, -0.0077], + [-0.0043, -0.0018, -0.0001, ..., 0.0055, -0.0032, -0.0046], + [-0.0015, 0.0055, -0.0029, ..., 0.0037, 0.0025, -0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.output_proj.weight': tensor([[ 8.4229e-03, -3.6133e-02, -1.4099e-02, ..., -6.5918e-03, + 5.1270e-03, 5.0659e-03], + [ 6.6223e-03, -7.2327e-03, 5.8594e-03, ..., -2.6703e-03, + 1.8311e-03, 3.2043e-03], + [-3.2196e-03, 5.2214e-05, 7.5684e-03, ..., 4.5776e-03, + 4.5166e-03, 2.8839e-03], + ..., + [ 2.7008e-03, 5.2490e-03, 7.0801e-03, ..., -4.4250e-03, + -6.4697e-03, -6.5002e-03], + [-2.2125e-03, 5.1270e-03, -6.3477e-03, ..., -9.4604e-03, + 8.1787e-03, 4.3640e-03], + [-4.2114e-03, 1.0132e-02, 3.9368e-03, ..., 4.6997e-03, + 4.4861e-03, 4.4556e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_proj.weight': tensor([[ 0.0085, 0.0082, -0.0034, ..., -0.0069, 0.0062, -0.0014], + [ 0.0073, -0.0261, -0.0047, ..., 0.0066, -0.0082, 0.0026], + [-0.0121, 0.0070, -0.0002, ..., 0.0057, 0.0209, -0.0048], + ..., + [ 0.0033, 0.0048, -0.0010, ..., -0.0012, -0.0046, 0.0108], + [-0.0066, -0.0033, 0.0006, ..., 0.0024, 0.0063, -0.0033], + [-0.0024, -0.0065, 0.0020, ..., -0.0033, -0.0007, 0.0029]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_norm.weight': tensor([ 9.7275e-04, 1.0315e-02, 3.7107e-10, 2.0752e-03, 2.0294e-03, + 2.3723e-05, -1.9550e-05, 5.0964e-03, 3.4180e-02, 2.6001e-02, + 6.8918e-07, 2.4536e-02, 5.5469e-01, 7.4768e-03, 8.4229e-03, + 5.3711e-03, 4.6875e-01, 1.6689e-05, 3.6163e-03, 1.1265e-05, + 5.1498e-04, 4.2725e-03, 5.9814e-02, 1.5991e-02, 1.1963e-02, + 6.3171e-03, 2.5146e-02, -7.5817e-05, 9.9487e-03, 6.7444e-03, + 4.1602e-01, 2.5146e-02, 4.8256e-04, 4.2188e-01, 5.3125e-01, + 1.9684e-03, 3.4790e-03, 3.4668e-02, 1.1475e-01, 1.4893e-02, + 3.8818e-02, 4.0894e-03, 6.4028e-09, 1.4496e-03, 1.8921e-02, + 2.6855e-02, 1.1719e-02, 9.7656e-03, 1.4954e-03, 9.8348e-06, + 8.5449e-03, -9.7156e-06, -7.1526e-06, 1.6809e-05, -4.6539e-04, + -2.6941e-05, 1.8677e-02, 1.0620e-02, 1.0620e-02, 2.9907e-03, + 4.0054e-04, 1.8716e-05, 8.7280e-03, -1.3530e-05, 1.3828e-04, + 1.4782e-04, 1.7047e-05, 2.5391e-01, 7.4387e-04, 1.7090e-02, + 4.0430e-01, -1.9455e-03, 3.9307e-02, 4.9744e-03, 5.4240e-06, + 1.3351e-04, 9.6436e-03, 4.9591e-04, 2.9449e-03, 1.8188e-02, + 2.0386e-02, 1.6968e-02, 1.9043e-01, 6.1951e-03, 3.4094e-05, + 1.5564e-02, 5.4321e-03, 2.0862e-05, 6.1512e-05, 4.7656e-01, + 7.6599e-03, -8.3447e-05, 1.0193e-02, 3.0100e-06, -5.2691e-05, + 4.3555e-01, 2.8419e-04, 7.9346e-03, -1.8304e-11, 1.0834e-03, + 4.1797e-01, 1.7166e-05, 9.5825e-03, -4.8578e-06, 8.3618e-03, + -3.4571e-05, 2.0409e-04, -2.4033e-04, 2.3804e-02, 1.1597e-02, + 6.1951e-03, -4.7982e-06, 3.1494e-02, 1.1301e-04, 1.1841e-02, + 1.4114e-04, 2.5513e-02, 4.1504e-03, 6.7749e-03, 4.1199e-04, + 2.5272e-05, 1.1169e-02, 8.7280e-03, 6.9885e-03, 1.5198e-02, + 2.6512e-04, 2.6855e-03, 5.3167e-05], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_proj.weight': tensor([[-0.0008, 0.0007, 0.0026, ..., -0.0004, -0.0014, 0.0005], + [-0.0063, -0.0021, -0.0051, ..., -0.0028, -0.0053, -0.0034], + [-0.0005, -0.0015, -0.0016, ..., -0.0027, -0.0029, -0.0014], + ..., + [ 0.0026, -0.0025, 0.0018, ..., 0.0011, 0.0024, -0.0006], + [ 0.0010, 0.0005, -0.0018, ..., -0.0025, 0.0009, 0.0019], + [ 0.0035, 0.0019, -0.0023, ..., -0.0041, -0.0006, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_norm.weight': tensor([-9.7275e-04, 1.0254e-02, 3.6016e-10, 1.8921e-03, 2.0142e-03, + 2.7061e-05, 1.9550e-05, 5.0659e-03, 3.3203e-02, 2.6123e-02, + -7.1898e-07, 2.4902e-02, 5.5859e-01, 7.4463e-03, 7.9346e-03, + 5.3406e-03, 4.6875e-01, 1.7524e-05, 3.5248e-03, -2.5392e-05, + 5.7983e-04, 4.3030e-03, 5.9814e-02, 1.5991e-02, 1.1780e-02, + 6.3782e-03, 2.5024e-02, -7.5817e-05, 9.3384e-03, 6.7444e-03, + 4.1602e-01, 2.5024e-02, 4.5395e-04, 4.2188e-01, 5.3125e-01, + 1.8234e-03, 3.5553e-03, 3.4912e-02, 1.1572e-01, 1.4893e-02, + 3.8818e-02, 4.1199e-03, -6.2282e-09, 1.2131e-03, 1.8555e-02, + 2.6733e-02, 1.1719e-02, 9.7656e-03, 1.3580e-03, 9.8944e-06, + 8.4839e-03, -1.0431e-05, -7.1526e-06, 1.9789e-05, 4.6539e-04, + 1.3947e-05, 1.8433e-02, 1.0742e-02, 1.0925e-02, 2.9907e-03, + -2.8419e-04, 1.8716e-05, 8.4229e-03, -1.3888e-05, 8.6308e-05, + 1.8692e-04, 1.6928e-05, 2.5391e-01, 6.4468e-04, 1.7090e-02, + 4.0430e-01, -1.9455e-03, 3.7842e-02, 4.9438e-03, -1.7956e-06, + 1.6117e-04, 8.6670e-03, 4.7112e-04, 2.9297e-03, 1.8188e-02, + 2.0264e-02, 1.7090e-02, 1.9043e-01, 6.0425e-03, 2.6345e-05, + 1.5564e-02, 5.8289e-03, 2.2173e-05, 5.7459e-05, 4.7656e-01, + 7.6599e-03, -9.9182e-05, 9.8877e-03, 2.2948e-06, -5.4359e-05, + 4.3555e-01, -1.0967e-04, 7.9956e-03, 1.8304e-11, 1.0605e-03, + 4.1797e-01, 1.6928e-05, 1.0498e-02, -7.1526e-06, 8.3618e-03, + -3.5048e-05, 2.2697e-04, -2.2984e-04, 2.3438e-02, 1.1536e-02, + 6.2866e-03, -4.2617e-06, 3.1128e-02, 1.1396e-04, 1.2024e-02, + -1.0920e-04, 2.5513e-02, 4.2725e-03, 7.2021e-03, 4.2534e-04, + 2.5153e-05, 1.1292e-02, 8.7280e-03, 6.9885e-03, 1.4771e-02, + 1.6975e-04, 2.7466e-03, -4.8637e-05], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.v_proj.weight': tensor([[-0.0104, 0.0016, 0.0097, ..., 0.0032, 0.0028, -0.0110], + [-0.0243, -0.0013, 0.0063, ..., -0.0010, -0.0076, 0.0008], + [-0.0132, 0.0032, 0.0044, ..., 0.0031, 0.0153, -0.0128], + ..., + [ 0.0117, 0.0008, 0.0018, ..., -0.0023, 0.0064, 0.0181], + [-0.0046, -0.0002, 0.0117, ..., -0.0018, -0.0161, -0.0034], + [-0.0216, -0.0033, 0.0044, ..., -0.0035, -0.0064, -0.0221]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.output_proj.weight': tensor([[ 0.0051, -0.0060, -0.0038, ..., -0.0053, -0.0006, 0.0098], + [ 0.0043, -0.0039, -0.0088, ..., 0.0214, -0.0043, -0.0114], + [-0.0044, 0.0049, 0.0037, ..., 0.0083, 0.0037, 0.0073], + ..., + [-0.0089, -0.0048, 0.0079, ..., -0.0076, -0.0035, -0.0005], + [ 0.0046, -0.0064, 0.0063, ..., 0.0055, -0.0054, -0.0113], + [ 0.0150, -0.0101, -0.0053, ..., 0.0051, -0.0061, 0.0074]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_img.weight': tensor([[ 0.0053, 0.0117, 0.0039, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0072, 0.0249, 0.0291], + [-0.0095, -0.0042, -0.0115, ..., -0.0342, -0.0327, 0.0066], + ..., + [ 0.0078, 0.0045, 0.0045, ..., 0.0086, 0.0286, -0.0165], + [ 0.0068, -0.0053, -0.0101, ..., -0.0101, 0.0142, -0.0309], + [ 0.0242, -0.0102, 0.0255, ..., -0.0227, -0.0044, -0.0153]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, 0.0067], + [ 0.0042, 0.0121, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0084, ..., -0.0059, -0.0057, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0037, 0.0049], + [-0.0160, 0.0051, 0.0039, ..., 0.0133, -0.0173, -0.0046], + [-0.0051, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0098]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0048, -0.0364, ..., -0.0225, -0.0142, -0.0056], + [-0.0157, 0.0051, -0.0364, ..., 0.0101, 0.0120, 0.0036], + [ 0.0043, 0.0615, 0.0010, ..., -0.0205, -0.0306, -0.0115], + ..., + [-0.0038, -0.0038, -0.0103, ..., -0.0211, -0.0019, 0.0288], + [ 0.0266, -0.0123, -0.0044, ..., -0.0420, -0.0282, 0.0053], + [ 0.0286, 0.0325, 0.0093, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.mlp.layer1.weight': tensor([[ 0.0073, 0.0172, -0.0126, ..., 0.0037, -0.0081, 0.0063], + [-0.0317, 0.0056, -0.0040, ..., -0.0054, 0.0044, -0.0088], + [-0.0048, -0.0020, 0.0048, ..., 0.0035, 0.0062, -0.0051], + ..., + [-0.0073, -0.0065, -0.0038, ..., -0.0095, 0.0120, 0.0034], + [ 0.0161, 0.0081, 0.0046, ..., -0.0063, 0.0014, 0.0034], + [-0.0178, -0.0041, -0.0038, ..., 0.0040, -0.0090, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.mlp.layer2.weight': tensor([[ 0.0001, -0.0159, 0.0052, ..., -0.0053, 0.0208, -0.0045], + [-0.0054, -0.0058, -0.0045, ..., -0.0073, 0.0044, -0.0053], + [ 0.0075, 0.0032, 0.0086, ..., 0.0044, 0.0024, -0.0157], + ..., + [ 0.0098, -0.0069, -0.0067, ..., 0.0011, 0.0050, -0.0056], + [-0.0063, 0.0007, 0.0046, ..., 0.0077, 0.0239, -0.0121], + [ 0.0059, -0.0151, 0.0050, ..., 0.0119, -0.0114, -0.0085]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0576, -0.0562, -0.0547, ..., 0.0029, 0.0030, 0.0030], + [ 0.0559, 0.0549, 0.0532, ..., -0.0023, -0.0024, -0.0025], + [ 0.0554, 0.0544, 0.0532, ..., -0.0025, -0.0025, -0.0026], + ..., + [ 0.0566, 0.0557, 0.0535, ..., -0.0022, -0.0024, -0.0024], + [ 0.0535, 0.0525, 0.0505, ..., -0.0021, -0.0021, -0.0021], + [ 0.0571, 0.0559, 0.0547, ..., -0.0029, -0.0030, -0.0029]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_self_attn.2.weight': tensor([[ 0.0019, -0.0016, -0.0020, ..., -0.0028, -0.0024, -0.0018], + [ 0.0017, -0.0021, -0.0018, ..., 0.0024, -0.0017, -0.0018], + [ 0.0002, 0.0033, 0.0002, ..., 0.0035, 0.0034, -0.0003], + ..., + [-0.0374, 0.0452, 0.0442, ..., 0.0420, 0.0374, 0.0374], + [-0.0100, 0.0103, 0.0104, ..., 0.0090, 0.0061, 0.0103], + [ 0.0078, -0.0095, -0.0092, ..., -0.0069, -0.0066, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_cross_attn.1.weight': tensor([[-9.0942e-03, -8.6670e-03, 7.8125e-03, ..., -6.0654e-04, + -5.3406e-04, -7.0572e-04], + [-4.1992e-02, -4.0527e-02, -4.0039e-02, ..., 2.9945e-04, + 3.0518e-04, 3.1471e-04], + [ 4.3213e-02, 4.2236e-02, 4.1748e-02, ..., -8.7738e-05, + -6.7711e-05, -9.2506e-05], + ..., + [ 4.1748e-02, 4.0771e-02, 4.0283e-02, ..., -3.7766e-04, + -3.1090e-04, -3.4714e-04], + [ 5.4688e-02, 5.3223e-02, 5.2002e-02, ..., -2.0981e-04, + -1.3924e-04, -1.9932e-04], + [ 3.8574e-02, 3.6865e-02, 3.6377e-02, ..., -4.7386e-06, + 8.0585e-05, 3.0994e-05]], dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0184, -0.0074, 0.0074, ..., 0.0074, 0.0109, 0.0067], + [-0.0044, -0.0042, 0.0042, ..., 0.0042, 0.0062, 0.0042], + [-0.0007, 0.0049, -0.0077, ..., -0.0049, -0.0074, -0.0078], + ..., + [-0.0078, -0.0265, 0.0332, ..., 0.0265, 0.0432, 0.0286], + [ 0.0114, 0.0063, 0.0122, ..., -0.0059, 0.0040, 0.0134], + [-0.0136, 0.0040, -0.0077, ..., -0.0040, -0.0046, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_mlp.1.weight': tensor([[-0.0037, -0.0033, -0.0031, ..., 0.0002, 0.0004, 0.0004], + [ 0.0540, 0.0518, 0.0503, ..., -0.0020, -0.0023, -0.0024], + [ 0.0459, 0.0444, 0.0437, ..., -0.0033, -0.0036, -0.0036], + ..., + [ 0.0884, 0.0854, 0.0845, ..., -0.0005, -0.0007, -0.0008], + [ 0.0035, 0.0035, 0.0036, ..., -0.0002, -0.0003, -0.0002], + [ 0.0654, 0.0635, 0.0618, ..., -0.0010, -0.0010, -0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_mlp.2.weight': tensor([[-0.0004, 0.0125, 0.0136, ..., 0.0120, 0.0003, 0.0098], + [-0.0039, -0.0009, 0.0010, ..., -0.0024, 0.0039, -0.0014], + [-0.0015, 0.0054, 0.0057, ..., 0.0036, 0.0015, 0.0044], + ..., + [ 0.0084, 0.0654, 0.0845, ..., 0.0688, -0.0084, 0.0500], + [ 0.0065, -0.0017, -0.0024, ..., -0.0022, -0.0064, -0.0020], + [-0.0120, -0.0071, -0.0072, ..., -0.0064, 0.0119, -0.0076]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.before_proj.weight': tensor([[ 0.0081, 0.0043, -0.0024, ..., -0.0027, -0.0028, 0.0038], + [-0.0078, -0.0134, -0.0053, ..., 0.0205, 0.0040, 0.0084], + [-0.0042, -0.0051, 0.0022, ..., 0.0023, 0.0043, -0.0065], + ..., + [-0.0031, 0.0089, 0.0085, ..., 0.0024, -0.0022, -0.0054], + [ 0.0143, 0.0198, 0.0146, ..., 0.0020, 0.0107, 0.0045], + [-0.0003, -0.0062, 0.0044, ..., 0.0002, -0.0022, -0.0056]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.before_proj.bias': tensor([ 0.0018, 0.0023, 0.0036, ..., 0.0046, 0.0033, -0.0115], + dtype=torch.bfloat16), 'net.control_blocks_2.0.after_proj.weight': tensor([[-0.0176, 0.0060, -0.0047, ..., -0.0038, -0.0049, -0.0137], + [-0.0106, -0.0037, -0.0068, ..., -0.0107, 0.0072, 0.0188], + [ 0.0019, 0.0009, -0.0058, ..., -0.0017, 0.0006, -0.0014], + ..., + [-0.0062, -0.0015, -0.0006, ..., 0.0024, 0.0011, -0.0017], + [ 0.0145, -0.0029, 0.0059, ..., -0.0047, -0.0109, 0.0077], + [ 0.0077, 0.0045, -0.0050, ..., 0.0039, -0.0055, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.after_proj.bias': tensor([ 0.0300, -0.0136, 0.0002, ..., -0.0012, -0.0262, 0.0337], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_proj.weight': tensor([[-0.0199, 0.0188, 0.0041, ..., 0.0259, -0.0085, 0.0050], + [-0.0106, -0.0162, 0.0114, ..., -0.0073, -0.0008, -0.0393], + [ 0.0139, 0.0078, 0.0051, ..., 0.0129, 0.0022, -0.0157], + ..., + [ 0.0159, 0.0104, -0.0030, ..., 0.0101, -0.0117, -0.0049], + [-0.0114, -0.0149, -0.0050, ..., -0.0016, -0.0167, -0.0042], + [-0.0027, -0.0081, -0.0071, ..., -0.0084, 0.0095, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_norm.weight': tensor([ 3.0664e-01, 9.6875e-01, 5.7129e-02, 7.6562e-01, 9.4922e-01, + -6.4319e-09, -3.0640e-02, 5.4123e-15, -4.0149e-19, 5.0093e-13, + 3.9916e-28, -8.3778e-33, -5.4919e-35, 3.3253e-22, -2.3074e-29, + 3.4907e-34, 1.4941e-01, 2.9325e-05, 1.5068e-04, -1.8921e-02, + 5.1880e-03, -1.3447e-04, 2.6367e-01, 9.4922e-01, 1.1875e+00, + 3.4766e-01, 7.3547e-03, 4.2383e-01, 1.7090e-02, 4.6327e-12, + 1.3281e-17, 4.6629e-15, 5.6839e-04, 1.4343e-03, -3.5095e-04, + 1.3184e-02, 1.4782e-04, 2.6375e-10, -3.8338e-04, 4.2848e-16, + 2.6455e-17, -2.0539e-14, -3.1896e-21, 8.8281e-01, 2.7148e-01, + 2.0117e-01, 4.7852e-01, 1.1406e+00, 1.2402e-01, -3.6926e-03, + -2.1118e-02, 2.0020e-01, 2.8992e-04, 1.8917e-10, -3.8147e-06, + 7.4593e-16, 7.2632e-03, 4.1034e-13, 7.5731e-29, -6.0486e-34, + 3.8556e-35, 1.6515e-15, -1.2494e-20, 1.0532e-33, 3.4766e-01, + 2.5586e-01, 5.8203e-01, 2.0996e-01, 1.1914e-01, -3.4925e-09, + 1.9287e-02, 2.2898e-15, -1.7364e-19, -5.2935e-13, -7.8097e-29, + 8.7269e-34, 2.7836e-35, -9.9262e-23, -1.1107e-27, -1.1059e-34, + 1.4191e-03, 2.7466e-04, 1.6403e-04, -1.2970e-04, 3.0708e-04, + 4.2236e-02, 9.4922e-01, 2.1582e-01, 2.0312e-01, 1.0781e+00, + 1.9824e-01, 1.8945e-01, -4.1504e-03, -4.0075e-12, -3.6592e-18, + 3.8303e-15, 1.0757e-03, 1.6724e-02, -3.8910e-03, -4.3640e-03, + 3.3188e-04, 5.5252e-11, 2.2799e-06, -7.3726e-18, 6.6058e-15, + -5.0848e-14, -5.5318e-24, 2.0312e-01, 1.1484e+00, 1.1953e+00, + 1.3750e+00, 4.0234e-01, 3.8281e-01, 9.8877e-03, -2.1484e-02, + -2.2583e-02, -7.4863e-05, 1.9554e-10, 2.4915e-05, 1.3791e-16, + 1.6846e-02, 1.0747e-13, 2.2404e-28, -6.4398e-34, -1.0391e-35, + -1.3214e-19, 7.1998e-20, 7.7037e-34], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.self_attn.k_proj.weight': tensor([[ 0.0152, 0.0017, -0.0059, ..., -0.0053, 0.0058, -0.0119], + [ 0.0231, 0.0071, 0.0044, ..., -0.0135, -0.0038, -0.0149], + [ 0.0060, 0.0164, -0.0024, ..., -0.0013, 0.0027, 0.0041], + ..., + [-0.0061, 0.0035, -0.0077, ..., 0.0115, 0.0005, -0.0031], + [-0.0100, -0.0090, -0.0032, ..., 0.0115, -0.0021, -0.0073], + [-0.0007, -0.0018, -0.0008, ..., -0.0059, 0.0046, 0.0048]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.k_norm.weight': tensor([ 3.2617e-01, 9.8047e-01, -1.0205e-01, 7.8516e-01, 9.4922e-01, + 2.9686e-09, 2.6245e-02, -5.9119e-15, -3.9980e-19, 5.4357e-13, + 4.5438e-28, -3.2259e-33, 4.4387e-35, 3.3418e-22, -2.1201e-29, + 3.4155e-34, 1.4941e-01, 2.9206e-05, -7.9155e-05, 5.4932e-02, + 5.1880e-03, 8.8215e-05, 2.9492e-01, 9.5312e-01, 1.1719e+00, + 1.4258e-01, 1.4844e-01, 4.6289e-01, 1.4404e-02, -4.5759e-12, + 1.3390e-17, 4.4964e-15, 5.3406e-04, -1.9302e-03, -3.2806e-04, + 4.2236e-02, 1.4687e-04, 2.1009e-10, -2.3651e-04, -4.2848e-16, + 2.4395e-17, -1.9540e-14, -3.1896e-21, 8.7891e-01, 2.5586e-01, + -4.7302e-03, -3.2227e-01, 1.1484e+00, 1.3672e-01, -3.0518e-03, + 2.8809e-02, 1.2402e-01, 1.0633e-04, -1.5280e-10, -3.4720e-06, + -7.4593e-16, 8.4229e-03, 4.1034e-13, 6.9814e-29, 9.3287e-35, + -1.2413e-35, -1.6515e-15, -1.1911e-20, -1.9410e-34, 3.8281e-01, + 3.1250e-01, 5.8594e-01, 1.6895e-01, -9.9609e-02, -4.1036e-09, + -2.2827e-02, -1.5543e-15, 4.5401e-19, -5.7909e-13, 2.1299e-29, + 1.0171e-33, 1.9184e-35, 8.2718e-23, 1.0539e-27, -4.8750e-34, + 3.4332e-04, 3.2663e-05, -1.6594e-04, -1.0920e-04, 3.0899e-04, + 4.2236e-02, 9.4531e-01, 2.0898e-01, 1.9824e-01, 1.0703e+00, + 1.8555e-01, 2.2266e-01, 9.3994e-03, 3.8085e-12, -4.4181e-18, + -3.6915e-15, 1.2054e-03, -1.5747e-02, -3.8910e-03, 3.7598e-02, + 3.2997e-04, 2.8922e-10, 1.6317e-06, -4.7976e-18, 6.6891e-15, + -5.0848e-14, 5.8678e-24, 2.9297e-01, 1.1406e+00, 1.1953e+00, + 1.3750e+00, 1.2891e-01, 3.9648e-01, -9.7656e-03, 4.3030e-03, + -1.6992e-01, 3.4332e-04, -1.7280e-10, -2.5034e-05, -1.3618e-16, + -1.5991e-02, -1.0569e-13, 1.9958e-28, -4.8750e-34, -3.7616e-35, + 3.3475e-18, 7.1574e-20, -4.5139e-34], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.self_attn.v_proj.weight': tensor([[ 8.2397e-03, 5.5847e-03, 7.1106e-03, ..., -2.0630e-02, + 8.9722e-03, 4.8828e-03], + [ 1.1108e-02, 6.4697e-03, 8.4686e-04, ..., -3.6163e-03, + -4.8523e-03, -4.3335e-03], + [-2.0905e-03, -3.7079e-03, 9.7275e-05, ..., -6.8359e-03, + 4.9744e-03, -7.9346e-03], + ..., + [ 3.6926e-03, 8.4229e-03, 1.3428e-02, ..., 2.0508e-02, + 3.6926e-03, -1.7578e-02], + [-1.8555e-02, 5.4321e-03, 1.1169e-02, ..., 3.5889e-02, + -7.6294e-03, 6.4850e-04], + [-5.1880e-03, -4.0588e-03, -5.6152e-03, ..., 1.9531e-02, + -5.6458e-03, 4.4861e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.output_proj.weight': tensor([[-0.0062, -0.0013, -0.0136, ..., 0.0078, -0.0044, -0.0070], + [ 0.0066, 0.0037, 0.0198, ..., 0.0023, -0.0028, 0.0106], + [-0.0044, -0.0047, -0.0019, ..., -0.0060, 0.0150, 0.0024], + ..., + [ 0.0079, 0.0010, -0.0066, ..., -0.0031, -0.0305, -0.0067], + [ 0.0038, -0.0014, -0.0115, ..., -0.0028, -0.0035, -0.0024], + [-0.0033, 0.0103, -0.0242, ..., 0.0129, 0.0017, -0.0036]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_proj.weight': tensor([[ 0.0050, 0.0261, 0.0086, ..., 0.0053, 0.0153, -0.0013], + [ 0.0194, 0.0076, 0.0065, ..., 0.0080, -0.0079, -0.0320], + [ 0.0176, 0.0128, 0.0025, ..., -0.0038, -0.0118, -0.0253], + ..., + [ 0.0061, 0.0032, -0.0038, ..., 0.0105, -0.0020, 0.0168], + [ 0.0058, 0.0090, 0.0104, ..., -0.0047, -0.0013, 0.0248], + [ 0.0175, -0.0170, -0.0084, ..., 0.0095, 0.0043, -0.0243]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_norm.weight': tensor([ 1.8799e-02, 4.7070e-01, 2.3560e-02, 1.0315e-02, 1.2107e-07, + 6.2937e-10, 3.8338e-04, 5.0391e-01, 8.6212e-04, 4.1602e-01, + 1.9165e-02, 1.9287e-02, 3.5095e-03, 1.3000e-02, 2.3315e-02, + 1.3977e-02, 2.6733e-02, -3.8147e-05, -4.0293e-05, 3.5400e-02, + 2.6367e-01, 3.2227e-02, 1.7700e-03, 1.4343e-02, 3.5400e-03, + 4.1008e-05, 1.8716e-05, 4.5410e-02, -6.2883e-06, 3.8281e-01, + 8.7738e-05, -1.3828e-04, 6.3171e-03, -2.5034e-05, 1.0864e-02, + 1.0193e-02, -2.0646e-10, 4.1406e-01, 1.0803e-02, 2.5635e-03, + 7.7148e-02, 2.8534e-03, 9.8877e-03, 1.0490e-04, 7.6904e-03, + 1.3977e-02, 2.9144e-03, 1.7212e-02, 1.4587e-02, 1.4709e-02, + 7.6904e-03, 4.0039e-01, 1.3611e-02, 1.1719e-02, 3.1494e-02, + 4.5538e-05, 9.1309e-02, 1.4709e-02, 1.5564e-02, 1.3672e-02, + 5.4836e-06, 7.6904e-03, 1.1597e-02, -1.4019e-04, 1.1414e-02, + 5.0659e-03, -8.0585e-05, 3.7689e-03, 1.8921e-02, 8.1543e-02, + 5.9128e-04, 6.1646e-03, 5.2979e-02, 4.1748e-02, 2.3956e-03, + 1.2695e-02, 4.0531e-06, 1.8677e-02, 3.1006e-02, 2.0996e-02, + 5.1880e-03, 1.6113e-02, 2.8320e-02, 3.6955e-06, 2.5391e-02, + 9.3994e-03, 3.5400e-03, 3.3203e-02, 6.9275e-03, 2.9602e-03, + 1.6724e-02, 2.0981e-04, 4.2419e-03, 2.1362e-02, 1.6724e-02, + -1.3351e-04, 9.9487e-03, 3.2227e-02, 1.5320e-02, 3.8086e-02, + 3.0762e-02, 2.4780e-02, 1.1169e-02, 6.7711e-05, -4.9360e-08, + 6.4392e-03, 1.8692e-04, 2.0264e-02, 8.1787e-03, 1.2573e-02, + 4.1809e-03, 1.1414e-02, 2.0508e-02, 5.1172e-01, 2.0447e-03, + 9.6436e-03, 1.0376e-02, 5.7068e-03, 7.8735e-03, 7.9590e-02, + 9.2983e-06, 4.5538e-05, 1.6451e-05, 1.2695e-02, 1.5259e-02, + 1.6357e-02, 2.4902e-02, 1.2573e-02], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_proj.weight': tensor([[ 0.0014, -0.0033, -0.0022, ..., -0.0038, 0.0019, -0.0005], + [-0.0085, -0.0035, -0.0021, ..., 0.0098, 0.0134, 0.0070], + [ 0.0055, 0.0036, 0.0045, ..., 0.0072, 0.0049, -0.0034], + ..., + [-0.0057, -0.0049, 0.0025, ..., -0.0029, -0.0009, -0.0032], + [-0.0078, -0.0053, -0.0126, ..., -0.0014, 0.0068, 0.0006], + [-0.0042, -0.0004, -0.0013, ..., 0.0044, -0.0016, -0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_norm.weight': tensor([ 1.9531e-02, 4.7070e-01, 2.3560e-02, 1.0437e-02, 1.2107e-07, + -2.2828e-10, 4.1962e-04, 5.0391e-01, 8.6594e-04, 4.1602e-01, + 2.0020e-02, 1.9287e-02, 3.5858e-03, 1.3000e-02, 2.3193e-02, + 1.3977e-02, 2.7100e-02, -3.8385e-05, -3.9816e-05, 3.5645e-02, + 2.6367e-01, 3.1982e-02, 1.7776e-03, 1.4587e-02, 3.6163e-03, + 4.0054e-05, 1.8597e-05, 4.5410e-02, -3.4124e-06, 3.8281e-01, + 8.7261e-05, -1.3924e-04, 6.1646e-03, -2.4796e-05, 1.0803e-02, + 1.0315e-02, -1.9008e-10, 4.1406e-01, 1.0864e-02, 2.4567e-03, + 7.6172e-02, 2.8839e-03, 9.6436e-03, 1.0633e-04, 7.5989e-03, + 1.3916e-02, 2.8687e-03, 1.7090e-02, 1.4404e-02, 1.5015e-02, + 7.5989e-03, 4.0039e-01, 1.3611e-02, 1.1230e-02, 3.1250e-02, + 4.4346e-05, 9.1309e-02, 1.4771e-02, 1.5259e-02, 1.3794e-02, + 6.3479e-06, 8.0566e-03, 1.1597e-02, -1.4019e-04, 1.1475e-02, + 4.9744e-03, -8.0585e-05, 3.7689e-03, 1.9043e-02, 8.1543e-02, + 5.6839e-04, 6.1340e-03, 5.2979e-02, 4.1748e-02, 2.4261e-03, + 1.2878e-02, 4.9472e-06, 1.8555e-02, 3.0884e-02, 2.1118e-02, + 5.1880e-03, 1.6113e-02, 2.7710e-02, 4.0531e-06, 2.5513e-02, + 9.3994e-03, 3.3875e-03, 3.2959e-02, 7.2632e-03, 3.0365e-03, + 1.6968e-02, 2.1172e-04, 4.3335e-03, 2.1240e-02, 1.6602e-02, + -1.3351e-04, 9.2773e-03, 3.1982e-02, 1.5259e-02, 3.7598e-02, + 3.1128e-02, 2.4780e-02, 1.1230e-02, 6.6280e-05, -1.7229e-08, + 6.4392e-03, 1.9360e-04, 1.9775e-02, 8.3008e-03, 1.2451e-02, + 4.2114e-03, 1.1169e-02, 2.0020e-02, 5.1172e-01, 2.0752e-03, + 9.5215e-03, 1.0315e-02, 5.5847e-03, 7.9956e-03, 7.9590e-02, + 1.3411e-05, 4.4584e-05, 1.6093e-05, 1.3000e-02, 1.5320e-02, + 1.6602e-02, 2.5024e-02, 1.2329e-02], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.v_proj.weight': tensor([[-0.0210, 0.0036, 0.0016, ..., 0.0012, -0.0206, -0.0220], + [-0.0035, 0.0081, 0.0038, ..., 0.0037, 0.0134, 0.0095], + [-0.0422, -0.0024, -0.0076, ..., -0.0079, -0.0286, -0.0074], + ..., + [-0.0376, 0.0016, -0.0150, ..., 0.0014, 0.0220, 0.0054], + [ 0.0082, 0.0015, -0.0062, ..., 0.0018, 0.0128, 0.0084], + [ 0.0112, 0.0050, -0.0072, ..., 0.0006, 0.0069, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.output_proj.weight': tensor([[ 7.2327e-03, 9.3384e-03, 4.6692e-03, ..., -6.8359e-03, + 2.5330e-03, 6.9885e-03], + [-3.9368e-03, -1.3184e-02, 6.2866e-03, ..., -8.1787e-03, + 6.6833e-03, 8.4229e-03], + [ 2.6245e-03, -2.0599e-03, -1.9684e-03, ..., -1.6785e-04, + -6.1417e-04, -7.0190e-03], + ..., + [-6.5918e-03, 6.3896e-05, 4.1809e-03, ..., 5.4626e-03, + 2.7466e-03, 7.8735e-03], + [ 7.6294e-03, 1.1658e-02, -5.0735e-04, ..., -2.6123e-02, + 7.1106e-03, 2.1606e-02], + [ 4.5166e-03, 4.0588e-03, -1.7456e-02, ..., -1.0742e-02, + 6.0120e-03, 7.8735e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_img.weight': tensor([[ 0.0079, 0.0034, 0.0430, ..., -0.0327, 0.0042, 0.0146], + [-0.0278, 0.0280, 0.0249, ..., -0.0046, -0.0378, 0.0271], + [ 0.0142, -0.0120, 0.0312, ..., -0.0074, 0.0253, -0.0118], + ..., + [ 0.0245, -0.0152, 0.0027, ..., -0.0452, 0.0216, -0.0209], + [-0.0071, 0.0074, -0.0471, ..., -0.0101, 0.0157, -0.0083], + [ 0.0085, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0028, -0.0122, -0.0115], + [ 0.0137, 0.0062, -0.0188, ..., 0.0157, 0.0082, -0.0047], + [-0.0306, -0.0051, -0.0273, ..., 0.0103, 0.0337, -0.0048], + ..., + [-0.0071, 0.0003, 0.0063, ..., 0.0209, -0.0106, 0.0081], + [ 0.0194, 0.0210, 0.0123, ..., -0.0146, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0109, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img.weight': tensor([[-0.0003, -0.0076, 0.0173, ..., 0.0084, 0.0356, 0.0217], + [-0.0498, -0.0049, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0104, -0.0013, -0.0139, ..., 0.0041, 0.0269, 0.0082], + ..., + [-0.0045, 0.0120, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0077, -0.0162, ..., -0.0059, -0.0076, -0.0121], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0134, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.mlp.layer1.weight': tensor([[-0.0031, -0.0219, -0.0079, ..., 0.0020, -0.0061, -0.0039], + [ 0.0006, 0.0044, 0.0055, ..., 0.0060, -0.0048, -0.0092], + [-0.0221, 0.0100, 0.0028, ..., -0.0045, -0.0090, -0.0073], + ..., + [ 0.0286, 0.0098, 0.0014, ..., 0.0140, 0.0062, 0.0143], + [-0.0159, -0.0058, 0.0041, ..., 0.0157, -0.0101, -0.0102], + [ 0.0052, -0.0114, -0.0042, ..., -0.0054, 0.0070, -0.0181]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.mlp.layer2.weight': tensor([[-0.0096, -0.0122, -0.0084, ..., -0.0148, -0.0040, -0.0079], + [ 0.0123, 0.0056, 0.0131, ..., 0.0037, 0.0097, -0.0077], + [ 0.0052, -0.0040, -0.0037, ..., -0.0010, -0.0015, 0.0038], + ..., + [ 0.0037, -0.0034, 0.0304, ..., -0.0075, -0.0130, 0.0051], + [ 0.0050, 0.0050, 0.0135, ..., 0.0075, 0.0049, -0.0062], + [ 0.0220, -0.0164, -0.0082, ..., 0.0068, 0.0042, 0.0048]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0432, -0.0420, -0.0415, ..., 0.0050, 0.0053, 0.0052], + [ 0.0352, 0.0339, 0.0330, ..., -0.0036, -0.0035, -0.0036], + [-0.0422, -0.0403, -0.0398, ..., 0.0050, 0.0051, 0.0050], + ..., + [-0.0476, -0.0464, -0.0449, ..., 0.0034, 0.0035, 0.0036], + [-0.0413, -0.0410, -0.0393, ..., 0.0036, 0.0039, 0.0039], + [-0.0251, -0.0240, -0.0228, ..., 0.0011, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0042, -0.0036, 0.0035, ..., 0.0033, 0.0033, 0.0034], + [ 0.0006, 0.0030, 0.0020, ..., 0.0028, 0.0019, -0.0021], + [-0.0068, 0.0038, -0.0041, ..., -0.0077, -0.0069, -0.0038], + ..., + [-0.0231, -0.0007, -0.0286, ..., -0.0308, -0.0292, 0.0007], + [-0.0117, -0.0054, -0.0129, ..., -0.0142, -0.0121, 0.0092], + [ 0.0069, -0.0063, 0.0044, ..., 0.0069, 0.0062, -0.0011]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_cross_attn.1.weight': tensor([[ 6.2500e-02, 6.0791e-02, 6.0303e-02, ..., -1.9550e-05, + -4.8876e-05, -4.3213e-06], + [ 7.8125e-03, 7.8125e-03, 7.8125e-03, ..., -3.9577e-05, + -1.1349e-04, -2.2888e-04], + [-2.0264e-02, -1.9531e-02, -1.8555e-02, ..., -6.7139e-04, + -6.7139e-04, -6.8283e-04], + ..., + [-2.6489e-02, -2.6489e-02, -2.5635e-02, ..., -3.9864e-04, + -3.5667e-04, -3.1853e-04], + [-2.7100e-02, -2.6489e-02, -2.5635e-02, ..., -2.0123e-04, + -1.5545e-04, -2.0885e-04], + [ 5.7373e-02, 5.5664e-02, 5.4932e-02, ..., -2.0409e-04, + -8.5831e-05, -2.3174e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0074, -0.0041, -0.0145, ..., 0.0043, 0.0080, -0.0095], + [ 0.0068, 0.0026, -0.0005, ..., -0.0011, -0.0033, 0.0017], + [ 0.0147, -0.0042, -0.0042, ..., 0.0043, -0.0070, 0.0095], + ..., + [ 0.0679, 0.0045, -0.0058, ..., 0.0059, -0.0175, 0.0630], + [ 0.0078, -0.0104, -0.0044, ..., -0.0155, -0.0220, 0.0183], + [-0.0065, 0.0087, -0.0153, ..., 0.0106, 0.0091, -0.0083]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_mlp.1.weight': tensor([[ 0.0096, 0.0090, 0.0082, ..., 0.0003, 0.0002, 0.0002], + [ 0.0325, 0.0317, 0.0315, ..., -0.0037, -0.0037, -0.0036], + [ 0.0457, 0.0444, 0.0439, ..., -0.0038, -0.0038, -0.0038], + ..., + [ 0.0085, 0.0082, 0.0088, ..., -0.0020, -0.0020, -0.0019], + [-0.0181, -0.0167, -0.0159, ..., 0.0015, 0.0014, 0.0014], + [ 0.0052, 0.0052, 0.0052, ..., -0.0024, -0.0025, -0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_mlp.2.weight': tensor([[-0.0024, -0.0014, -0.0033, ..., 0.0035, 0.0031, 0.0024], + [-0.0004, -0.0010, 0.0036, ..., 0.0014, 0.0003, 0.0015], + [ 0.0036, -0.0046, -0.0056, ..., -0.0038, -0.0011, -0.0038], + ..., + [-0.0011, 0.0083, 0.0131, ..., -0.0092, 0.0016, -0.0082], + [ 0.0167, 0.0129, 0.0238, ..., -0.0120, -0.0175, -0.0125], + [ 0.0157, -0.0153, -0.0156, ..., -0.0150, -0.0161, -0.0140]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.after_proj.weight': tensor([[-0.0284, 0.0079, 0.0044, ..., -0.0072, -0.0132, 0.0128], + [ 0.0086, 0.0190, 0.0042, ..., -0.0061, -0.0064, 0.0031], + [ 0.0120, -0.0156, -0.0128, ..., -0.0083, 0.0104, 0.0078], + ..., + [-0.0187, -0.0171, 0.0166, ..., -0.0065, -0.0040, -0.0048], + [-0.0098, -0.0085, 0.0060, ..., 0.0056, 0.0134, -0.0042], + [-0.0182, 0.0089, -0.0125, ..., -0.0096, -0.0066, -0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.after_proj.bias': tensor([ 0.0125, 0.0034, 0.0019, ..., 0.0067, 0.0037, -0.0056], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_proj.weight': tensor([[ 0.0243, -0.0233, -0.0165, ..., 0.0219, -0.0100, -0.0254], + [-0.0001, -0.0242, -0.0025, ..., -0.0129, 0.0074, 0.0037], + [ 0.0148, 0.0154, 0.0070, ..., 0.0044, -0.0095, 0.0009], + ..., + [-0.0036, -0.0067, -0.0034, ..., 0.0040, 0.0106, -0.0075], + [ 0.0281, 0.0144, -0.0022, ..., 0.0077, -0.0160, 0.0281], + [ 0.0074, -0.0193, 0.0089, ..., -0.0330, 0.0123, -0.0231]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_norm.weight': tensor([ 1.5625e+00, 4.9414e-01, 5.1562e-01, 3.0273e-01, 1.4922e+00, + 4.3335e-03, -1.1719e-02, -4.3392e-05, 1.8616e-03, 8.9722e-03, + 1.4827e-06, 3.5623e-08, 1.2266e+00, -6.4392e-03, -3.1876e-17, + -3.2715e-02, -1.3367e-02, -4.3213e-02, 2.4414e-03, 1.8188e-02, + 1.6895e-01, 2.8372e-05, 6.4062e-01, 1.5859e+00, 6.0547e-01, + 6.8750e-01, 4.7656e-01, 1.6699e-01, 2.0020e-02, 5.5847e-03, + 2.7710e-02, -4.9591e-05, 6.1340e-03, -2.0874e-02, 3.3691e-02, + 4.4434e-02, 1.2451e-01, 8.9722e-03, -7.6599e-03, 1.5736e-05, + -5.0545e-05, -2.2507e-04, 1.2146e-02, 1.3125e+00, 1.3184e-01, + 1.3906e+00, 6.6797e-01, 2.1094e-01, -3.7598e-02, -1.5869e-02, + 2.9297e-02, 5.1025e-02, 1.3477e-01, 3.3855e-05, 2.4080e-05, + -5.6252e-07, 4.5061e-05, -7.8678e-05, -6.1279e-02, -4.2783e-09, + 4.6875e-02, -1.9932e-04, -3.3691e-02, -3.5889e-02, 8.7500e-01, + 1.6016e+00, 1.4688e+00, 1.4531e+00, 3.1250e-01, -4.6997e-03, + 5.7068e-03, 3.9339e-05, 1.2939e-02, -4.2114e-03, -1.9932e-04, + 9.0525e-07, 3.2715e-02, 2.4780e-02, 7.6050e-15, 8.1250e-01, + 1.0078e+00, -4.6387e-03, -1.1368e-03, 3.2959e-02, 1.2988e-01, + 2.0508e-02, 1.5234e+00, 5.6641e-01, 1.3984e+00, 1.5000e+00, + 2.8125e-01, 5.1562e-01, 5.1025e-02, -1.1902e-02, 2.1606e-02, + -1.8024e-04, -2.0752e-03, -3.7231e-03, 1.8463e-03, 3.2959e-02, + 1.2207e-01, 2.5177e-04, 1.1641e+00, 1.7405e-05, 2.4915e-05, + -4.7684e-05, -8.5354e-05, 9.6875e-01, 1.5391e+00, 1.4844e+00, + 1.5469e+00, 3.7891e-01, 3.2227e-02, 4.5410e-02, 2.2827e-02, + 1.9287e-02, 1.4746e-01, 6.7711e-05, -7.6890e-06, 1.7658e-06, + 1.8120e-05, 9.2030e-05, 1.1094e+00, 2.7831e-10, 2.4536e-02, + -1.2219e-05, 7.6562e-01, 1.1172e+00], dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.self_attn.k_proj.weight': tensor([[-0.0388, 0.0049, -0.0071, ..., 0.0249, 0.0139, 0.0238], + [-0.0056, 0.0130, -0.0036, ..., -0.0114, 0.0283, -0.0051], + [-0.0294, 0.0200, -0.0027, ..., 0.0057, 0.0228, -0.0034], + ..., + [-0.0057, 0.0018, -0.0045, ..., 0.0026, 0.0011, 0.0127], + [ 0.0050, 0.0120, -0.0009, ..., -0.0046, -0.0048, 0.0089], + [ 0.0190, -0.0225, -0.0019, ..., -0.0131, 0.0058, -0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.k_norm.weight': tensor([ 1.6328e+00, 8.9844e-01, 3.9648e-01, 3.6328e-01, 1.4922e+00, + 4.3030e-03, 1.6235e-02, -4.3392e-05, 6.6528e-03, 8.9722e-03, + 3.5018e-06, -1.6065e-08, 1.2266e+00, -6.0425e-03, -8.1966e-17, + -2.6489e-02, -4.1199e-03, 1.1768e-01, 4.2725e-03, 1.9836e-03, + 2.7466e-02, -7.7724e-05, 6.5625e-01, 1.4297e+00, 2.5391e-01, + 7.0312e-01, 5.5859e-01, 2.3926e-01, -1.8677e-02, 1.3000e-02, + 2.9663e-02, 1.3947e-05, 6.5308e-03, 5.4688e-02, 3.4912e-02, + -1.8311e-02, -1.2573e-02, 8.9722e-03, 2.6489e-02, 1.5974e-05, + -8.2016e-05, -2.2030e-04, 1.2024e-02, 1.5078e+00, 9.9219e-01, + 1.3750e+00, 7.3047e-01, 2.5586e-01, 6.5308e-03, 6.7871e-02, + 3.1738e-02, 7.4158e-03, 3.6774e-03, 6.2466e-05, -1.4484e-05, + -4.8429e-07, 2.9206e-05, 4.9114e-05, 9.8877e-03, -4.5693e-09, + -2.2217e-02, -1.9741e-04, -4.4434e-02, -3.8818e-02, 7.5781e-01, + 1.4609e+00, 1.5000e+00, 1.4375e+00, 2.9688e-01, 4.4556e-03, + -1.6937e-03, 3.6716e-05, 2.1515e-03, 4.3030e-03, -1.9932e-04, + -8.9034e-07, -3.0029e-02, 2.5024e-02, 7.4940e-15, 8.1250e-01, + 1.0078e+00, 5.7617e-02, 1.6022e-03, 1.1719e-02, -3.0884e-02, + 2.0752e-02, 1.4844e+00, 9.2969e-01, 1.5156e+00, 1.5391e+00, + 4.8584e-02, 5.1172e-01, 9.9487e-03, 6.8970e-03, 1.5625e-02, + 2.0885e-04, -6.2180e-04, 4.5166e-02, 4.4250e-03, 1.0498e-02, + 1.1414e-02, -1.1492e-04, 1.1641e+00, 1.7643e-05, 2.4319e-05, + -4.7207e-05, -4.1246e-05, 6.7188e-01, 1.3672e+00, 1.3828e+00, + 1.6641e+00, 3.8086e-01, -4.3701e-02, 1.1963e-01, 2.4170e-02, + 2.2583e-02, -1.5259e-02, -1.0729e-06, -2.0266e-06, -1.4156e-06, + 4.1485e-05, -1.6212e-04, 1.1094e+00, -3.0923e-10, -1.7853e-03, + -8.9407e-06, 7.6562e-01, 1.1172e+00], dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.self_attn.v_proj.weight': tensor([[ 0.0025, -0.0052, -0.0096, ..., -0.0293, -0.0051, 0.0109], + [ 0.0068, 0.0021, -0.0122, ..., -0.0017, -0.0008, -0.0042], + [-0.0245, 0.0041, -0.0022, ..., 0.0148, 0.0055, -0.0143], + ..., + [ 0.0137, 0.0065, -0.0337, ..., -0.0454, 0.0104, -0.0117], + [ 0.0017, -0.0108, 0.0018, ..., -0.0069, -0.0074, -0.0117], + [-0.0015, 0.0028, -0.0188, ..., -0.0042, 0.0068, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.output_proj.weight': tensor([[-0.0040, 0.0315, -0.0271, ..., -0.0048, -0.0188, 0.0134], + [ 0.0038, -0.0175, -0.0282, ..., 0.0054, 0.0064, -0.0032], + [-0.0034, -0.0038, 0.0041, ..., 0.0038, -0.0100, 0.0074], + ..., + [ 0.0094, -0.0041, -0.0128, ..., 0.0277, 0.0123, -0.0049], + [ 0.0029, 0.0190, -0.0044, ..., 0.0171, 0.0071, -0.0211], + [-0.0037, -0.0123, 0.0038, ..., -0.0239, 0.0061, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_proj.weight': tensor([[ 0.0055, -0.0037, 0.0042, ..., -0.0042, 0.0129, 0.0038], + [ 0.0061, 0.0057, 0.0023, ..., 0.0026, 0.0038, -0.0031], + [ 0.0036, -0.0056, 0.0003, ..., -0.0069, 0.0186, -0.0042], + ..., + [-0.0025, 0.0046, 0.0038, ..., 0.0108, 0.0037, 0.0046], + [-0.0059, -0.0049, 0.0035, ..., 0.0124, 0.0095, 0.0079], + [ 0.0080, 0.0013, -0.0031, ..., -0.0109, 0.0103, -0.0040]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_norm.weight': tensor([ 6.7749e-03, 1.2305e-01, 2.2339e-02, 6.0156e-01, 2.3315e-02, + 9.6436e-03, 2.0386e-02, 4.0588e-03, 1.5137e-02, 1.0645e-01, + 2.2583e-02, 4.7852e-02, 2.6172e-01, 2.6489e-02, 1.7969e-01, + 1.8433e-02, 4.6875e-02, 2.4048e-02, 5.5847e-03, 8.3008e-03, + 1.2268e-02, 1.5918e-01, 2.7148e-01, 1.9238e-01, 2.1118e-02, + 1.4941e-01, 4.4922e-02, 8.5449e-03, 4.0820e-01, 1.3611e-02, + 6.3782e-03, 5.0391e-01, 4.4922e-02, 4.1748e-02, 2.0752e-02, + 8.9722e-03, 7.5684e-02, 5.2344e-01, 5.9375e-01, 3.5400e-03, + 1.3062e-02, 1.0156e-01, 8.3618e-03, 5.5859e-01, 1.6479e-02, + 3.2715e-02, 8.4839e-03, 4.2773e-01, 4.2419e-03, 2.1484e-02, + 6.3477e-02, 2.2949e-02, 1.7334e-02, 5.6641e-01, -7.0095e-05, + 3.2227e-02, 3.3691e-02, 6.7383e-02, 1.5198e-02, 1.5198e-02, + 2.5330e-03, 1.2329e-02, 1.5717e-03, 3.5547e-01, 1.4648e-02, + -1.2684e-04, 1.8433e-02, 5.8350e-02, 7.4707e-02, 9.2163e-03, + 9.2773e-03, 6.3672e-01, 1.9789e-05, 6.0938e-01, 1.5381e-02, + 6.7139e-04, 9.5825e-03, 2.9297e-01, 2.9907e-02, 2.9564e-04, + 9.3262e-02, 8.3008e-03, 4.1748e-02, 7.6294e-03, 2.4170e-02, + 5.2490e-03, 2.2827e-02, 6.4941e-02, 2.0898e-01, 9.8877e-03, + 5.9814e-03, 1.3489e-02, 4.0527e-02, 2.0752e-02, 3.5156e-02, + 2.5269e-02, 9.3384e-03, 2.4414e-02, 5.0293e-02, 1.9531e-02, + 2.9907e-02, 1.5625e-02, -3.8300e-03, 5.1514e-02, 6.1719e-01, + 5.2734e-02, 1.1749e-03, 4.9219e-01, 1.8799e-02, 2.3315e-02, + 1.1902e-02, 2.0294e-03, -6.1512e-05, 8.0078e-02, 4.6094e-01, + 1.4099e-02, 2.3682e-02, 6.1328e-01, 1.1084e-01, 3.0518e-02, + 1.1169e-02, 6.9922e-01, 1.4453e-01, 1.3916e-02, 7.8125e-03, + 2.4414e-03, 1.8921e-02, 1.1520e-03], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_proj.weight': tensor([[ 0.0038, -0.0034, 0.0050, ..., -0.0018, -0.0112, 0.0114], + [ 0.0039, 0.0003, -0.0025, ..., 0.0014, -0.0126, -0.0041], + [ 0.0024, -0.0016, -0.0035, ..., 0.0120, -0.0132, -0.0042], + ..., + [ 0.0052, 0.0046, -0.0034, ..., 0.0023, 0.0061, -0.0125], + [ 0.0093, 0.0029, -0.0020, ..., -0.0042, 0.0042, 0.0027], + [-0.0091, 0.0007, -0.0101, ..., -0.0042, -0.0160, 0.0046]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_norm.weight': tensor([ 6.7749e-03, 1.2305e-01, 2.2461e-02, 6.0156e-01, 2.3193e-02, + 9.4604e-03, 2.0386e-02, 4.0588e-03, 1.5381e-02, 1.0645e-01, + 2.2583e-02, 4.6143e-02, 2.6172e-01, 2.6611e-02, 1.7969e-01, + 1.9409e-02, 4.6387e-02, 2.4048e-02, 5.5237e-03, 8.1787e-03, + 1.2390e-02, 1.6016e-01, 2.7344e-01, 1.9238e-01, 2.0752e-02, + 1.4941e-01, 4.3945e-02, 6.3477e-03, 4.1016e-01, 1.3489e-02, + 6.6528e-03, 5.0391e-01, 4.4922e-02, 4.1992e-02, 2.1118e-02, + 8.3008e-03, 7.4707e-02, 5.2344e-01, 5.9375e-01, 3.6011e-03, + 1.3062e-02, 1.0107e-01, 8.3618e-03, 5.5859e-01, 1.6846e-02, + 3.2959e-02, 8.7891e-03, 4.2773e-01, 4.3335e-03, 2.1362e-02, + 6.3477e-02, 2.3193e-02, 1.7334e-02, 5.6641e-01, -7.0095e-05, + 3.2227e-02, 3.3203e-02, 6.7383e-02, 1.5259e-02, 1.3672e-02, + 2.4261e-03, 1.2268e-02, 2.3956e-03, 3.5547e-01, 1.4221e-02, + 2.9755e-04, 1.7822e-02, 5.8350e-02, 7.4707e-02, 9.3384e-03, + 9.5825e-03, 6.3672e-01, 1.9789e-05, 6.0938e-01, 1.5625e-02, + -5.8365e-04, 9.6436e-03, 2.9297e-01, 3.0640e-02, 3.8910e-04, + 9.3750e-02, 8.1177e-03, 4.0771e-02, 7.3547e-03, 2.3682e-02, + 5.2490e-03, 2.3193e-02, 6.5430e-02, 2.0898e-01, 9.8877e-03, + 6.0120e-03, 1.2329e-02, 4.0527e-02, 2.0996e-02, 3.6133e-02, + 2.5269e-02, 9.3994e-03, 2.4414e-02, 5.0293e-02, 1.9409e-02, + 2.9175e-02, 1.5747e-02, 3.8300e-03, 5.2246e-02, 6.1719e-01, + 5.2734e-02, 1.1444e-03, 4.9219e-01, 1.8433e-02, 2.3071e-02, + 1.2695e-02, 2.0294e-03, -5.4836e-05, 8.0078e-02, 4.6094e-01, + 1.3672e-02, 2.3438e-02, 6.1328e-01, 1.1182e-01, 3.0640e-02, + 1.0864e-02, 7.0312e-01, 1.4355e-01, 1.2756e-02, 7.5378e-03, + 2.3346e-03, 1.8677e-02, -9.3460e-04], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.v_proj.weight': tensor([[ 0.0133, -0.0066, -0.0176, ..., -0.0031, -0.0162, -0.0116], + [-0.0157, 0.0014, -0.0052, ..., -0.0026, -0.0210, 0.0210], + [-0.0116, -0.0044, -0.0012, ..., -0.0032, 0.0299, 0.0249], + ..., + [-0.0052, 0.0014, -0.0125, ..., 0.0035, 0.0112, -0.0006], + [ 0.0284, 0.0068, 0.0254, ..., 0.0093, 0.0096, 0.0153], + [ 0.0162, 0.0038, 0.0074, ..., 0.0071, -0.0186, -0.0244]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.output_proj.weight': tensor([[ 0.0153, 0.0042, -0.0161, ..., 0.0133, 0.0081, 0.0150], + [ 0.0088, -0.0010, -0.0029, ..., 0.0084, 0.0030, 0.0031], + [-0.0039, -0.0069, 0.0044, ..., -0.0048, 0.0067, -0.0067], + ..., + [ 0.0047, -0.0082, 0.0043, ..., 0.0047, -0.0096, 0.0056], + [ 0.0045, -0.0133, 0.0067, ..., -0.0403, -0.0249, -0.0233], + [-0.0021, -0.0025, 0.0154, ..., 0.0086, 0.0036, 0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0068, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [ 0.0038, 0.0150, -0.0159, ..., -0.0034, -0.0056, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0068, -0.0057], + [-0.0337, -0.0031, 0.0048, ..., -0.0032, 0.0172, 0.0231], + [-0.0255, 0.0004, -0.0044, ..., 0.0408, 0.0430, -0.0120]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0033], + [ 0.0198, -0.0208, 0.0129, ..., 0.0089, -0.0337, -0.0051], + [ 0.0302, 0.0101, -0.0018, ..., -0.0304, 0.0199, -0.0039], + ..., + [ 0.0171, -0.0276, -0.0066, ..., 0.0231, 0.0182, 0.0011], + [ 0.0221, 0.0388, 0.0047, ..., 0.0042, -0.0488, -0.0361], + [ 0.0045, 0.0118, -0.0059, ..., 0.0010, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img.weight': tensor([[ 0.0273, -0.0041, 0.0118, ..., 0.0022, 0.0347, 0.0615], + [-0.0258, -0.0047, -0.0197, ..., 0.0139, -0.0042, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0097, 0.0132, 0.0184, ..., 0.0110, 0.0089, 0.0515], + [-0.0017, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.mlp.layer1.weight': tensor([[-0.0108, 0.0063, 0.0072, ..., -0.0044, -0.0053, -0.0164], + [-0.0096, 0.0017, -0.0006, ..., -0.0088, 0.0035, -0.0081], + [-0.0038, 0.0062, 0.0038, ..., -0.0022, -0.0123, -0.0510], + ..., + [-0.0118, 0.0111, 0.0038, ..., 0.0068, -0.0043, -0.0075], + [-0.0062, 0.0075, 0.0042, ..., -0.0066, -0.0088, -0.0383], + [-0.0030, 0.0275, -0.0038, ..., 0.0016, 0.0046, -0.0094]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.mlp.layer2.weight': tensor([[-0.0107, -0.0052, -0.0210, ..., 0.0140, -0.0079, -0.0112], + [-0.0051, 0.0047, -0.0121, ..., -0.0032, 0.0078, -0.0229], + [-0.0051, 0.0137, -0.0172, ..., 0.0005, 0.0127, -0.0106], + ..., + [-0.0009, 0.0081, -0.0381, ..., 0.0082, 0.0142, -0.0155], + [-0.0127, 0.0184, 0.0122, ..., 0.0067, 0.0096, -0.0037], + [-0.0101, -0.0040, -0.0364, ..., -0.0003, 0.0109, -0.0334]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0289, 0.0291, 0.0275, ..., -0.0041, -0.0042, -0.0043], + [-0.0254, -0.0244, -0.0242, ..., 0.0043, 0.0047, 0.0046], + [-0.0297, -0.0293, -0.0286, ..., 0.0045, 0.0048, 0.0048], + ..., + [-0.0322, -0.0312, -0.0305, ..., 0.0021, 0.0022, 0.0023], + [-0.0327, -0.0325, -0.0317, ..., 0.0020, 0.0021, 0.0021], + [-0.0046, -0.0046, -0.0045, ..., 0.0001, 0.0001, 0.0001]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0049, -0.0042, -0.0014, ..., -0.0048, -0.0051, -0.0014], + [ 0.0017, -0.0020, -0.0017, ..., -0.0018, -0.0022, 0.0031], + [ 0.0099, -0.0096, -0.0102, ..., -0.0041, -0.0034, 0.0007], + ..., + [ 0.0388, -0.0427, -0.0432, ..., -0.0249, -0.0217, 0.0142], + [-0.0041, 0.0042, -0.0049, ..., 0.0044, 0.0045, 0.0060], + [-0.0148, 0.0173, 0.0148, ..., 0.0162, 0.0146, 0.0165]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_cross_attn.1.weight': tensor([[ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., -1.3065e-04, + -6.0320e-05, -2.1267e-04], + [ 6.1768e-02, 6.0303e-02, 5.8594e-02, ..., 3.3569e-04, + 4.4632e-04, 4.7874e-04], + [-5.3955e-02, -5.2490e-02, -5.1025e-02, ..., -5.2154e-06, + 5.0545e-05, -5.7459e-05], + ..., + [ 3.3447e-02, 3.2715e-02, 3.1982e-02, ..., 1.0204e-04, + 3.3617e-05, 6.2943e-05], + [ 5.9570e-02, 5.8105e-02, 5.7617e-02, ..., 5.2261e-04, + 5.6458e-04, 5.5695e-04], + [-4.8828e-02, -4.8096e-02, -4.7363e-02, ..., -3.2997e-04, + -3.7384e-04, -3.3379e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0254, 0.0063, -0.0156, ..., 0.0030, -0.0043, -0.0016], + [-0.0216, 0.0137, 0.0175, ..., -0.0090, -0.0137, 0.0099], + [ 0.0098, 0.0008, -0.0087, ..., 0.0060, 0.0008, -0.0008], + ..., + [ 0.0225, 0.0156, -0.0128, ..., 0.0072, 0.0080, 0.0045], + [ 0.0117, 0.0444, -0.0057, ..., 0.0040, 0.0225, -0.0079], + [-0.0187, 0.0161, 0.0192, ..., -0.0135, -0.0082, 0.0210]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_mlp.1.weight': tensor([[-0.0571, -0.0557, -0.0542, ..., 0.0029, 0.0031, 0.0031], + [-0.0259, -0.0244, -0.0232, ..., 0.0008, 0.0008, 0.0007], + [ 0.0344, 0.0332, 0.0332, ..., -0.0016, -0.0016, -0.0016], + ..., + [ 0.0304, 0.0294, 0.0295, ..., -0.0018, -0.0017, -0.0015], + [ 0.0356, 0.0342, 0.0327, ..., -0.0010, -0.0009, -0.0010], + [-0.0295, -0.0288, -0.0282, ..., 0.0021, 0.0021, 0.0020]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_mlp.2.weight': tensor([[-0.0233, -0.0145, 0.0118, ..., 0.0041, 0.0162, -0.0124], + [-0.0010, -0.0058, -0.0026, ..., -0.0016, 0.0057, 0.0014], + [-0.0004, -0.0009, 0.0005, ..., -0.0008, 0.0005, 0.0001], + ..., + [ 0.0432, -0.0041, -0.0190, ..., -0.0137, 0.0040, 0.0197], + [ 0.0226, -0.0126, -0.0173, ..., -0.0151, 0.0124, 0.0167], + [ 0.0072, -0.0306, -0.0190, ..., -0.0220, 0.0315, 0.0182]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.after_proj.weight': tensor([[-0.0034, 0.0076, -0.0066, ..., -0.0079, 0.0073, 0.0118], + [ 0.0095, -0.0077, 0.0047, ..., -0.0018, 0.0079, 0.0058], + [ 0.0027, 0.0077, 0.0167, ..., 0.0063, 0.0069, 0.0176], + ..., + [ 0.0086, -0.0040, 0.0145, ..., 0.0044, -0.0072, 0.0212], + [ 0.0033, -0.0074, 0.0148, ..., -0.0079, -0.0046, -0.0078], + [-0.0034, 0.0047, -0.0098, ..., -0.0081, 0.0216, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.after_proj.bias': tensor([-0.0038, -0.0012, 0.0049, ..., -0.0038, 0.0101, -0.0013], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_proj.weight': tensor([[-0.0121, 0.0339, 0.0119, ..., -0.0087, 0.0035, -0.0254], + [-0.0052, 0.0046, 0.0090, ..., 0.0026, 0.0165, -0.0212], + [ 0.0130, 0.0161, 0.0101, ..., -0.0052, -0.0282, 0.0151], + ..., + [-0.0036, -0.0046, 0.0041, ..., 0.0019, 0.0104, 0.0014], + [-0.0120, -0.0302, 0.0251, ..., -0.0388, 0.0052, -0.0098], + [-0.0039, -0.0037, 0.0004, ..., 0.0075, 0.0177, 0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_norm.weight': tensor([ 9.3750e-01, 1.7422e+00, 1.4844e+00, 1.5547e+00, 1.5430e-01, + 1.2207e-01, 1.7212e-02, -1.6065e-08, -2.3954e-33, 3.5553e-03, + 1.1353e-02, 1.0791e-01, -5.8594e-03, -4.5166e-03, 1.6022e-03, + -1.0490e-04, 6.5430e-02, 1.0391e+00, 1.1094e+00, 7.7820e-04, + 4.1485e-05, 1.2891e+00, 2.0312e+00, 1.8516e+00, 1.0625e+00, + 8.8281e-01, 1.3750e+00, -1.2817e-02, -6.9046e-04, 3.4714e-04, + -2.3041e-03, -2.5940e-04, 3.7354e-02, 1.1641e+00, 6.1035e-03, + 1.1172e+00, 1.6113e-02, 3.1710e-05, 1.4221e-02, -7.5378e-03, + 2.0862e-05, 1.6212e-05, 1.5020e-05, 2.0938e+00, 1.3516e+00, + 1.4922e+00, 1.5234e+00, 1.4688e+00, 9.4238e-02, 5.4932e-04, + 3.1982e-02, 2.7148e-01, 1.1719e+00, 1.1475e-01, 2.1362e-03, + 8.2520e-02, 2.8839e-03, 1.0107e-01, -4.4107e-05, 3.6163e-03, + -1.1978e-03, 5.2185e-03, -5.2490e-03, 1.1719e+00, 1.9375e+00, + 6.9531e-01, 8.5547e-01, 1.3828e+00, 4.5410e-02, -2.3193e-02, + 1.1475e-02, 3.3993e-08, -3.8037e-33, 9.4604e-03, 1.8921e-03, + -3.1586e-03, -1.5450e-04, 2.0386e-02, -4.1748e-02, 2.5635e-03, + 4.6387e-02, 4.7607e-03, 1.8188e-02, 8.7738e-04, 9.6321e-05, + -5.2185e-03, 8.4375e-01, 8.2031e-01, 1.5312e+00, 1.4609e+00, + 7.6562e-01, -5.9814e-03, -2.0885e-04, -5.9605e-05, -2.3651e-03, + 2.6894e-04, 2.1973e-02, 1.3770e-01, 2.3804e-03, 4.5654e-02, + -2.1362e-02, -8.4639e-06, 1.2891e-01, 6.4850e-05, 5.7983e-03, + -1.5736e-05, -2.9206e-05, 5.1953e-01, 1.7266e+00, 1.4922e+00, + 8.3984e-01, 7.5781e-01, 9.4727e-02, 4.0054e-04, 2.5391e-02, + 1.1562e+00, 8.0078e-02, 1.1875e+00, -7.3910e-05, 9.0332e-03, + 1.9043e-02, 4.7852e-02, 2.4438e-05, -2.1210e-03, 2.1648e-04, + -2.1729e-02, 1.1328e+00, -9.9487e-03], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.self_attn.k_proj.weight': tensor([[-1.5747e-02, 6.3171e-03, 2.8381e-03, ..., -1.1230e-02, + -3.0212e-03, 3.7842e-03], + [-1.0864e-02, -2.5635e-02, 1.5503e-02, ..., -7.3242e-03, + -4.9438e-03, -5.3406e-03], + [ 5.0964e-03, 4.3945e-03, 9.0942e-03, ..., 1.7090e-02, + -6.8665e-03, 3.7384e-03], + ..., + [-4.6082e-03, 4.9133e-03, 4.7913e-03, ..., 1.9836e-03, + -3.8605e-03, 3.3417e-03], + [-1.3550e-02, -1.5736e-05, 2.5757e-02, ..., -1.5381e-02, + 7.1716e-03, 5.3406e-03], + [ 1.9226e-03, 4.5166e-03, -7.6904e-03, ..., -4.6692e-03, + 6.8665e-03, 2.1362e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.k_norm.weight': tensor([ 8.7891e-01, 1.7109e+00, 1.5000e+00, 1.4922e+00, 1.5625e-01, + 1.2793e-01, 1.1902e-02, -6.7521e-09, -1.7935e-33, -1.6251e-03, + 1.1169e-02, 1.0889e-01, -1.0925e-02, 3.1006e-02, -2.0142e-03, + 9.8705e-05, 1.3123e-02, 1.0391e+00, 1.1094e+00, 7.6675e-04, + 7.2479e-05, 1.2891e+00, 1.7969e+00, 1.7188e+00, 1.2891e-01, + 3.8281e-01, 1.4219e+00, -1.5015e-02, 3.2425e-04, -3.0136e-04, + -2.4567e-03, 2.9564e-04, -3.6133e-02, 1.1641e+00, 2.5787e-03, + 1.1094e+00, 1.6113e-02, -3.0518e-05, -1.4954e-02, -7.6599e-03, + -1.0788e-05, 2.3365e-05, -5.0306e-05, 1.6016e+00, 6.2109e-01, + 1.4531e+00, 1.5781e+00, 1.4922e+00, -5.6885e-02, 5.0735e-04, + 1.9653e-02, -9.3994e-03, 1.1719e+00, 1.2354e-01, 2.1515e-03, + 8.2520e-02, -3.9368e-03, 1.0156e-01, -3.3855e-05, 7.0801e-02, + -1.2054e-03, 2.3315e-02, -1.5869e-02, 1.1719e+00, 1.9609e+00, + 9.9219e-01, 7.5781e-01, 1.3672e+00, 2.6855e-02, -1.0193e-02, + -2.4170e-02, 9.8255e-08, 4.4778e-33, 1.0010e-02, 1.4587e-02, + 4.3945e-02, 1.6403e-04, 3.7842e-02, -5.2490e-02, -2.6093e-03, + -4.6692e-03, 1.9287e-02, 1.5198e-02, 8.6975e-04, 9.9182e-05, + 3.2043e-03, 1.3828e+00, 1.3281e+00, 1.6250e+00, 1.5156e+00, + 3.1445e-01, 2.5177e-03, -6.5231e-04, -1.4687e-04, 2.4109e-03, + 3.3951e-04, 4.9744e-03, -2.4170e-02, -1.9684e-03, 7.1411e-03, + 2.1484e-02, -3.4332e-05, 1.2793e-01, -2.4080e-05, 5.7678e-03, + 1.9908e-05, -2.6584e-05, 1.4219e+00, 1.9531e+00, 1.4609e+00, + 1.3086e-01, 7.0703e-01, 5.0659e-03, 4.2915e-04, 4.6082e-03, + 1.1797e+00, -4.9438e-03, 1.1875e+00, 5.9366e-05, -1.8555e-02, + 1.9409e-02, 4.6387e-02, 3.0398e-05, 4.4189e-02, 2.1935e-04, + 2.8198e-02, 1.1328e+00, -1.0071e-02], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.self_attn.v_proj.weight': tensor([[-0.0090, -0.0089, 0.0096, ..., 0.0508, -0.0099, -0.0027], + [-0.0032, 0.0112, -0.0091, ..., 0.0052, -0.0070, -0.0080], + [-0.0040, 0.0182, -0.0139, ..., -0.0025, 0.0050, 0.0128], + ..., + [ 0.0008, 0.0008, -0.0097, ..., -0.0038, 0.0085, 0.0115], + [ 0.0040, 0.0105, -0.0109, ..., 0.0032, -0.0037, -0.0028], + [-0.0021, -0.0089, 0.0081, ..., 0.0151, 0.0078, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.output_proj.weight': tensor([[ 0.0198, 0.0126, 0.0085, ..., 0.0065, 0.0132, -0.0049], + [ 0.0198, -0.0013, 0.0054, ..., -0.0305, -0.0049, 0.0111], + [-0.0098, 0.0184, -0.0212, ..., -0.0133, -0.0063, 0.0136], + ..., + [-0.0292, -0.0199, 0.0059, ..., -0.0051, -0.0112, 0.0065], + [-0.0072, 0.0178, -0.0261, ..., -0.0131, 0.0439, 0.0035], + [-0.0305, -0.0042, 0.0276, ..., 0.0226, -0.0085, -0.0272]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_proj.weight': tensor([[-0.0057, -0.0057, 0.0170, ..., -0.0219, -0.0225, 0.0112], + [-0.0082, 0.0065, 0.0289, ..., -0.0229, -0.0029, 0.0105], + [ 0.0126, -0.0037, -0.0076, ..., 0.0052, 0.0110, -0.0043], + ..., + [ 0.0104, 0.0056, -0.0078, ..., -0.0189, -0.0057, 0.0044], + [ 0.0064, 0.0035, -0.0112, ..., 0.0045, -0.0034, 0.0016], + [-0.0064, -0.0037, 0.0124, ..., 0.0123, -0.0010, 0.0044]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_norm.weight': tensor([0.0967, 0.0156, 0.4414, 0.0576, 0.1416, 0.0320, 0.1377, 0.1201, 0.5195, + 0.0297, 0.0488, 0.0168, 0.0378, 0.0713, 0.0508, 0.1836, 0.4746, 0.0219, + 0.7578, 0.0405, 0.0283, 0.1533, 0.0262, 0.0757, 0.0425, 0.3906, 0.5156, + 0.2617, 0.0684, 0.0405, 0.0231, 0.2178, 0.6797, 0.2402, 0.0322, 0.0349, + 0.0141, 0.0928, 0.0175, 0.0359, 0.0491, 0.0986, 0.0206, 0.0298, 0.2041, + 0.0337, 0.3828, 0.0153, 0.0413, 0.0170, 0.0173, 0.4180, 0.4668, 0.3418, + 0.0708, 0.6211, 0.5977, 0.0447, 0.1060, 0.0198, 0.0767, 0.4648, 0.0071, + 0.5000, 0.1475, 0.3926, 0.0815, 0.0471, 0.0630, 0.4473, 0.0280, 0.4707, + 0.0135, 0.0366, 0.1377, 0.0698, 0.0148, 0.0991, 0.6523, 0.0337, 0.0732, + 0.0908, 0.0388, 0.0503, 0.4492, 0.0601, 0.0640, 0.0879, 0.0189, 0.1396, + 0.0327, 0.1758, 0.0408, 0.0549, 0.0376, 0.0996, 0.0088, 0.3340, 0.5195, + 0.0344, 0.0520, 0.2949, 0.0322, 0.1108, 0.0325, 0.2158, 0.0123, 0.6406, + 0.6211, 0.0175, 0.0161, 0.7188, 0.0410, 0.0160, 0.4805, 0.2090, 0.2695, + 0.6992, 0.0332, 0.0762, 0.0664, 0.5586, 0.6133, 0.0234, 0.0225, 0.1855, + 0.0366, 0.0369], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_proj.weight': tensor([[-3.1586e-03, 1.8158e-03, 4.1504e-03, ..., -1.4114e-03, + 1.0681e-02, 3.1281e-03], + [ 2.8076e-03, -6.7711e-05, -1.6861e-03, ..., 4.7913e-03, + -1.1108e-02, 5.2795e-03], + [-1.9043e-02, 7.6904e-03, -3.2043e-04, ..., 1.8433e-02, + 1.0681e-02, 3.5553e-03], + ..., + [ 5.8594e-03, 3.1128e-03, -5.7983e-03, ..., 9.4986e-04, + -1.9287e-02, 5.4932e-04], + [-9.5825e-03, -1.8539e-03, -1.0681e-02, ..., 6.0425e-03, + -1.0071e-02, 5.9509e-03], + [ 2.0264e-02, -1.4038e-03, 1.0681e-02, ..., -8.6670e-03, + 7.0190e-03, -6.2866e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_norm.weight': tensor([0.0967, 0.0157, 0.4414, 0.0574, 0.1416, 0.0327, 0.1387, 0.1206, 0.5195, + 0.0302, 0.0498, 0.0168, 0.0378, 0.0713, 0.0508, 0.1855, 0.4746, 0.0217, + 0.7539, 0.0403, 0.0281, 0.1533, 0.0264, 0.0752, 0.0422, 0.3867, 0.5156, + 0.2598, 0.0688, 0.0408, 0.0238, 0.2197, 0.6797, 0.2402, 0.0330, 0.0349, + 0.0137, 0.0928, 0.0175, 0.0354, 0.0483, 0.0996, 0.0210, 0.0303, 0.2051, + 0.0337, 0.3828, 0.0154, 0.0410, 0.0166, 0.0179, 0.4180, 0.4668, 0.3418, + 0.0708, 0.6211, 0.6016, 0.0439, 0.1050, 0.0201, 0.0767, 0.4648, 0.0075, + 0.5000, 0.1465, 0.3926, 0.0815, 0.0471, 0.0630, 0.4473, 0.0282, 0.4707, + 0.0132, 0.0366, 0.1377, 0.0693, 0.0148, 0.0996, 0.6523, 0.0334, 0.0723, + 0.0908, 0.0391, 0.0503, 0.4492, 0.0603, 0.0630, 0.0874, 0.0194, 0.1416, + 0.0337, 0.1748, 0.0408, 0.0552, 0.0378, 0.1006, 0.0087, 0.3340, 0.5195, + 0.0337, 0.0525, 0.2949, 0.0325, 0.1108, 0.0325, 0.2158, 0.0122, 0.6406, + 0.6211, 0.0172, 0.0162, 0.7188, 0.0408, 0.0161, 0.4824, 0.2090, 0.2695, + 0.6953, 0.0327, 0.0747, 0.0664, 0.5586, 0.6133, 0.0233, 0.0225, 0.1855, + 0.0378, 0.0374], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.v_proj.weight': tensor([[ 0.0226, 0.0038, -0.0243, ..., 0.0069, -0.0026, -0.0189], + [-0.0203, 0.0140, 0.0005, ..., 0.0066, -0.0025, -0.0311], + [ 0.0251, -0.0085, -0.0001, ..., -0.0010, -0.0099, -0.0452], + ..., + [ 0.0023, -0.0017, 0.0232, ..., -0.0073, -0.0212, 0.0034], + [-0.0327, -0.0032, -0.0256, ..., -0.0060, 0.0122, -0.0121], + [-0.0164, 0.0014, 0.0045, ..., -0.0046, -0.0042, 0.0089]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.output_proj.weight': tensor([[ 0.0017, 0.0038, -0.0017, ..., -0.0120, 0.0074, 0.0096], + [-0.0315, -0.0042, -0.0132, ..., -0.0078, -0.0199, -0.0077], + [ 0.0161, -0.0069, 0.0054, ..., -0.0117, 0.0118, -0.0172], + ..., + [ 0.0181, -0.0146, 0.0189, ..., 0.0119, -0.0060, -0.0081], + [-0.0137, -0.0216, 0.0060, ..., 0.0129, -0.0192, 0.0261], + [ 0.0005, -0.0215, 0.0315, ..., -0.0364, 0.0332, 0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0057, -0.0065, ..., 0.0266, -0.0278, 0.0075], + ..., + [-0.0217, -0.0410, 0.0488, ..., -0.0039, -0.0003, -0.0255], + [ 0.0036, 0.0066, 0.0013, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0059, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0101, 0.0038, ..., -0.0236, -0.0109, -0.0248], + [-0.0079, -0.0013, 0.0293, ..., 0.0072, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0143, 0.0126, -0.0094], + ..., + [-0.0035, 0.0320, -0.0233, ..., -0.0060, -0.0015, 0.0188], + [ 0.0136, -0.0352, 0.0267, ..., -0.0305, -0.0045, 0.0037], + [-0.0205, 0.0140, 0.0025, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img.weight': tensor([[-0.0079, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0029, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0057, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0018], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, -0.0027, -0.0020], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0023, ..., -0.0103, 0.0074, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.mlp.layer1.weight': tensor([[ 0.0151, -0.0038, 0.0084, ..., 0.0068, 0.0133, -0.0104], + [-0.0098, 0.0107, 0.0242, ..., 0.0029, 0.0115, -0.0012], + [-0.0037, 0.0354, -0.0305, ..., 0.0278, 0.0125, 0.0047], + ..., + [-0.0060, 0.0233, 0.0693, ..., -0.0088, 0.0012, -0.0090], + [-0.0055, -0.0082, -0.0050, ..., -0.0095, -0.0016, -0.0164], + [-0.0005, 0.0047, -0.0097, ..., -0.0092, -0.0284, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.mlp.layer2.weight': tensor([[ 0.0067, 0.0117, 0.0032, ..., -0.0220, -0.0221, -0.0071], + [ 0.0104, 0.0247, 0.0068, ..., -0.0092, 0.0356, -0.0181], + [ 0.0040, -0.0050, -0.0037, ..., 0.0150, 0.0275, 0.0042], + ..., + [ 0.0045, 0.0088, 0.0064, ..., -0.0083, -0.0062, 0.0048], + [ 0.0051, -0.0092, -0.0031, ..., 0.0077, 0.0183, 0.0037], + [-0.0036, 0.0048, 0.0170, ..., 0.0052, 0.0349, 0.0073]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0226, -0.0222, -0.0220, ..., 0.0010, 0.0008, 0.0009], + [-0.0315, -0.0306, -0.0298, ..., 0.0026, 0.0027, 0.0027], + [-0.0222, -0.0214, -0.0205, ..., 0.0021, 0.0023, 0.0023], + ..., + [ 0.0605, 0.0588, 0.0569, ..., -0.0037, -0.0037, -0.0037], + [ 0.0496, 0.0479, 0.0464, ..., -0.0032, -0.0031, -0.0030], + [-0.0256, -0.0250, -0.0248, ..., 0.0022, 0.0023, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0061, 0.0044, 0.0062, ..., 0.0099, 0.0130, 0.0059], + [ 0.0005, 0.0014, 0.0015, ..., -0.0015, 0.0004, 0.0016], + [ 0.0015, -0.0013, 0.0005, ..., 0.0037, 0.0038, -0.0014], + ..., + [-0.0237, -0.0762, -0.0615, ..., 0.0369, 0.0354, -0.0693], + [ 0.0089, 0.0082, 0.0085, ..., 0.0182, 0.0150, 0.0092], + [ 0.0164, 0.0099, 0.0124, ..., 0.0187, 0.0208, 0.0113]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_cross_attn.1.weight': tensor([[-8.2031e-02, -8.0078e-02, -7.7637e-02, ..., -1.0443e-04, + -1.2016e-04, -1.0967e-04], + [ 3.0029e-02, 2.9419e-02, 2.8564e-02, ..., 5.6624e-06, + 3.1948e-05, 1.5163e-04], + [ 3.2715e-02, 3.1494e-02, 3.0518e-02, ..., 2.6703e-04, + 2.8992e-04, 1.9836e-04], + ..., + [ 5.5908e-02, 5.4199e-02, 5.3223e-02, ..., -1.8978e-04, + -1.8501e-04, -1.1683e-04], + [ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.8311e-04, + -1.2112e-04, -1.7071e-04], + [-3.0884e-02, -2.9907e-02, -2.9175e-02, ..., -5.8889e-05, + 9.3460e-05, 1.1492e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_cross_attn.2.weight': tensor([[-7.9346e-03, 3.7842e-03, -8.8501e-03, ..., 2.8320e-02, + 3.0396e-02, 3.6926e-03], + [ 6.7139e-03, -9.3994e-03, 1.1841e-02, ..., -2.4605e-04, + 4.5166e-03, 4.6387e-03], + [ 2.0905e-03, -1.2207e-03, 4.6997e-03, ..., -7.9346e-03, + -6.3477e-03, -4.7913e-03], + ..., + [-1.5991e-02, 4.6387e-03, 3.5400e-02, ..., 2.5146e-02, + 3.2227e-02, -8.7891e-03], + [-2.3315e-02, 1.2085e-02, 3.6865e-02, ..., 1.0254e-02, + 1.5869e-02, -1.4099e-02], + [ 1.3977e-02, -8.5449e-03, -7.0801e-03, ..., 1.5717e-03, + 1.6098e-03, -4.5300e-05]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_mlp.1.weight': tensor([[-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., 1.3638e-04, + 2.5940e-04, 2.0313e-04], + [ 1.3245e-02, 1.1841e-02, 1.1597e-02, ..., 3.7193e-04, + 4.5204e-04, 4.4441e-04], + [-7.3242e-02, -7.1777e-02, -7.0801e-02, ..., -4.4823e-04, + -3.7766e-04, -4.1199e-04], + ..., + [-6.5430e-02, -6.4453e-02, -6.3477e-02, ..., 1.6785e-04, + 2.7084e-04, 1.6785e-04], + [-4.4922e-02, -4.4434e-02, -4.4189e-02, ..., 2.0981e-04, + 6.8188e-05, 7.9632e-05], + [ 6.2012e-02, 6.1035e-02, 6.0791e-02, ..., -1.4210e-04, + -1.6975e-04, -1.3351e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_mlp.2.weight': tensor([[-0.0141, 0.0030, -0.0100, ..., -0.0164, -0.0199, 0.0178], + [-0.0017, -0.0063, 0.0064, ..., -0.0015, 0.0049, -0.0028], + [-0.0016, -0.0029, -0.0036, ..., -0.0010, -0.0017, 0.0016], + ..., + [-0.0493, -0.0312, -0.0461, ..., -0.0361, -0.0109, 0.0266], + [ 0.0075, -0.0177, 0.0260, ..., 0.0075, 0.0052, -0.0075], + [ 0.0135, 0.0118, 0.0070, ..., 0.0121, 0.0072, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.after_proj.weight': tensor([[ 2.8229e-03, 1.6724e-02, -1.8311e-02, ..., 7.8125e-03, + 3.9368e-03, -6.8970e-03], + [-8.4229e-03, -3.6316e-03, 7.6599e-03, ..., -9.5844e-05, + 1.6235e-02, -8.6060e-03], + [ 9.6436e-03, 4.3335e-03, -3.9978e-03, ..., 3.8452e-03, + 2.1820e-03, -7.6599e-03], + ..., + [ 1.7456e-02, 1.2329e-02, -5.0354e-03, ..., 7.7515e-03, + -6.1035e-03, 5.2795e-03], + [-6.2866e-03, 7.4463e-03, -3.2349e-03, ..., 1.0437e-02, + -3.9673e-03, 4.2419e-03], + [ 3.4332e-03, -1.4832e-02, -5.6458e-03, ..., 2.1362e-02, + -1.2268e-02, 8.9722e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.after_proj.bias': tensor([ 0.0085, -0.0193, -0.0089, ..., -0.0041, -0.0183, 0.0126], + dtype=torch.bfloat16)} +[06-26 08:28:42|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:42|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 08:28:42|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 2 checkpoint. +[06-26 08:28:42|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/seg(5136ef49-6d8d-42e8-8abf-7dac722a304a) +[06-26 08:28:42|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 3 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/23057a4167b89de89a4a397fdbf3887994d115eb/general/seg/5136ef49-6d8d-42e8-8abf-7dac722a304a_ema_bf16.pt +[06-26 08:28:43|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0143, 0.0022, 0.0062], + [-0.0022, 0.0054, -0.0327, ..., -0.0080, 0.0010, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0134, -0.0026, 0.0283], + ..., + [-0.0026, 0.0125, 0.0248, ..., 0.0437, -0.0073, -0.0192], + [ 0.0277, 0.0148, 0.0054, ..., -0.0015, -0.0091, 0.0146], + [ 0.0437, 0.0366, 0.0082, ..., 0.0016, -0.0129, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0006, -0.0310, 0.0300, ..., -0.0067, -0.0043, -0.0339], + [-0.0104, -0.0066, 0.0111, ..., -0.0004, 0.0092, 0.0520], + [-0.0063, 0.0132, 0.0398, ..., -0.0101, -0.0179, -0.0064], + ..., + [-0.0034, 0.0175, -0.0096, ..., 0.0225, 0.0045, 0.0066], + [-0.0209, -0.0081, 0.0086, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0054, 0.0354, -0.0082, ..., 0.0117, 0.0356, -0.0186], + [ 0.0513, 0.0141, -0.0010, ..., 0.0376, 0.0532, 0.0026], + [ 0.0175, -0.0347, -0.0013, ..., -0.0063, 0.0057, -0.0283], + ..., + [ 0.0354, -0.0048, -0.0209, ..., -0.0153, 0.0262, -0.0113], + [ 0.0037, 0.0244, 0.0356, ..., -0.0026, 0.0170, 0.0117], + [ 0.0486, 0.0175, 0.0043, ..., 0.0016, -0.0033, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0117, ..., -0.0075, -0.0337, 0.0076], + [ 0.0491, 0.0413, 0.0251, ..., -0.0060, 0.0393, 0.0040], + [-0.0162, 0.0039, 0.0135, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0026, 0.0261, 0.0041, ..., -0.0038, 0.0347, 0.0053], + [-0.0254, 0.0068, -0.0012, ..., 0.0096, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0124, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0033, -0.0270, -0.0166, ..., 0.0117, -0.0298, 0.0532], + [-0.0016, 0.0044, -0.0129, ..., -0.0161, 0.0150, 0.0308], + [ 0.0137, 0.0084, 0.0139, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0053, -0.0630, 0.0005, ..., 0.0074, -0.0183, 0.0214], + [ 0.0034, -0.0018, 0.0258, ..., -0.0486, -0.0349, -0.0044], + [ 0.0182, -0.0074, 0.0198, ..., 0.0152, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0036, -0.0028, -0.0006, ..., 0.0035, 0.0022, -0.0117], + [-0.0154, 0.0251, 0.0432, ..., -0.0052, 0.0034, -0.0016], + [-0.0144, -0.0203, -0.0245, ..., -0.0112, -0.0325, -0.0018], + ..., + [ 0.0050, 0.0037, 0.0417, ..., 0.0281, 0.0025, 0.0030], + [-0.0038, -0.0046, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0227, -0.0364, -0.0220, ..., 0.0062, -0.0039, -0.0091]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0156, -0.0078, -0.0481, ..., 0.0449, 0.0212, 0.0127], + [ 0.0154, -0.0227, 0.0220, ..., -0.0221, -0.0020, 0.0086], + [-0.0151, -0.0095, 0.0430, ..., -0.0101, 0.0327, -0.0112], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0146], + [-0.0025, 0.0166, -0.0063, ..., 0.0002, 0.0081, -0.0078], + [-0.0112, -0.0015, 0.0123, ..., 0.0042, -0.0036, -0.0082]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0025, -0.0386, -0.0123, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0084, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0131, 0.0019, -0.0124, ..., 0.0214, 0.0105, 0.0050], + ..., + [-0.0153, -0.0172, -0.0140, ..., 0.0095, 0.0101, 0.0184], + [-0.0154, -0.0356, 0.0201, ..., -0.0047, -0.0114, 0.0332], + [-0.0396, -0.0119, -0.0092, ..., -0.0284, -0.0142, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0276, 0.0312, 0.0136, ..., -0.0195, 0.0112, -0.0142], + [ 0.0192, 0.0081, -0.0077, ..., 0.0018, -0.0135, -0.0198], + [-0.0248, 0.0100, -0.0139, ..., 0.0090, 0.0172, 0.0061], + ..., + [ 0.0466, -0.0047, 0.0003, ..., -0.0008, 0.0239, 0.0047], + [ 0.0547, -0.0194, -0.0106, ..., -0.0017, 0.0311, -0.0044], + [ 0.0035, 0.0101, -0.0098, ..., 0.0203, 0.0170, 0.0143]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0062, -0.0061, 0.0282, ..., -0.0111, -0.0038, -0.0182], + [ 0.0022, 0.0126, -0.0315, ..., -0.0457, -0.0079, 0.0105], + [-0.0009, 0.0261, 0.0245, ..., 0.0063, 0.0074, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0156, 0.0034, -0.0067], + [ 0.0076, -0.0371, 0.0356, ..., -0.0082, 0.0410, -0.0178], + [ 0.0178, -0.0146, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0078, 0.0208, 0.0276, ..., 0.0136, 0.0151, 0.0083], + [-0.0062, 0.0025, -0.0049, ..., 0.0145, 0.0005, 0.0048], + [ 0.0056, 0.0021, 0.0242, ..., -0.0120, -0.0070, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0059, ..., 0.0248, -0.0095, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., -0.0010, -0.0295, 0.0400], + [-0.0114, 0.0130, 0.0305, ..., -0.0107, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0049, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0146, -0.0043, 0.0288, ..., 0.0150, 0.0583, 0.0001], + [-0.0281, 0.0156, -0.0386, ..., -0.0444, -0.0093, -0.0013], + ..., + [ 0.0030, 0.0157, 0.0217, ..., -0.0005, 0.0115, 0.0151], + [ 0.0081, 0.0195, -0.0283, ..., -0.0009, 0.0021, -0.0342], + [-0.0162, -0.0195, -0.0013, ..., -0.0024, 0.0165, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0146, -0.0138, 0.0459, ..., -0.0018, -0.0320, 0.0278], + [ 0.0019, 0.0045, 0.0270, ..., -0.0220, -0.0183, -0.0271], + [ 0.0221, 0.0217, -0.0201, ..., 0.0030, 0.0062, -0.0053], + ..., + [-0.0258, 0.0034, 0.0215, ..., -0.0018, 0.0237, -0.0266], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0073], + [ 0.0026, -0.0082, -0.0130, ..., -0.0334, 0.0057, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0155, -0.0454, ..., -0.0134, -0.0128, 0.0400], + [ 0.0291, -0.0189, -0.0110, ..., -0.0288, 0.0025, -0.0186], + [ 0.0043, 0.0547, -0.0019, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0275, 0.0042, -0.0269, ..., -0.0096, -0.0287, 0.0150], + [ 0.0101, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0095, 0.0108, ..., -0.0058, 0.0703, -0.0052]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0091, -0.0315, ..., 0.0074, 0.0254, 0.0359], + [ 0.0079, 0.0051, 0.0040, ..., -0.0119, -0.0233, -0.0065], + [ 0.0046, -0.0037, -0.0118, ..., -0.0125, -0.0072, 0.0084], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, -0.0026], + [ 0.0339, -0.0066, 0.0223, ..., 0.0410, 0.0002, 0.0128], + [-0.0049, -0.0439, -0.0486, ..., 0.0137, -0.0129, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0053, 0.0131, -0.0023, ..., -0.0093, -0.0001, -0.0229], + [-0.0287, 0.0229, 0.0075, ..., -0.0137, 0.0045, 0.0222], + [ 0.0065, 0.0288, -0.0189, ..., -0.0173, 0.0311, -0.0262], + ..., + [ 0.0233, 0.0070, 0.0038, ..., -0.0023, 0.0278, 0.0064], + [-0.0109, 0.0005, 0.0151, ..., -0.0175, 0.0444, 0.0425], + [ 0.0062, -0.0068, 0.0093, ..., -0.0040, -0.0344, 0.0226]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[ 0.0059, 0.0125, 0.0190, ..., -0.0095, -0.0198, 0.0142], + [-0.0581, 0.0294, -0.0038, ..., -0.0240, 0.0066, 0.0302], + [ 0.0079, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0017, -0.0289, 0.0537, ..., 0.0065, -0.0056, -0.0093], + [ 0.0053, -0.0080, 0.0209, ..., -0.0035, 0.0090, 0.0186], + [-0.0082, -0.0349, -0.0173, ..., -0.0400, 0.0037, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0178, -0.0259, -0.0102, ..., 0.0090, -0.0137, -0.0405], + [ 0.0063, -0.0095, 0.0266, ..., -0.0400, 0.0047, 0.0391], + [ 0.0059, 0.0359, 0.0283, ..., 0.0295, -0.0233, 0.0212], + ..., + [ 0.0071, 0.0124, 0.0132, ..., -0.0139, -0.0033, 0.0095], + [ 0.0299, -0.0167, -0.0028, ..., -0.0052, -0.0032, -0.0206], + [ 0.0072, -0.0067, -0.0038, ..., -0.0032, -0.0047, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, 0.0056, 0.0116, ..., -0.0032, -0.0211, -0.0064], + [-0.0101, -0.0284, -0.0192, ..., 0.0376, 0.0210, 0.0229], + [ 0.0139, -0.0278, 0.0116, ..., 0.0396, -0.0061, -0.0069], + ..., + [ 0.0029, 0.0034, -0.0311, ..., 0.0344, 0.0062, -0.0125], + [-0.0236, -0.0104, -0.0061, ..., 0.0332, -0.0117, -0.0173], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0138, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[ 0.0014, 0.0166, 0.0006, ..., 0.0273, -0.0352, 0.0152], + [-0.0510, -0.0082, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0074, ..., -0.0388, 0.0004, 0.0398], + ..., + [-0.0160, -0.0195, -0.0021, ..., -0.0015, -0.0051, 0.0033], + [-0.0187, 0.0042, 0.0098, ..., 0.0287, -0.0120, -0.0101], + [-0.0181, 0.0039, 0.0247, ..., -0.0149, 0.0066, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0085, 0.0096, 0.0140, ..., -0.0112, -0.0023, -0.0253], + [-0.0166, 0.0330, -0.0171, ..., 0.0024, -0.0403, -0.0110], + [ 0.0086, 0.0347, 0.0275, ..., 0.0153, 0.0068, 0.0088], + ..., + [ 0.0271, -0.0168, -0.0105, ..., -0.0101, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0058, ..., -0.0040, 0.0006, -0.0243], + [-0.0083, 0.0581, 0.0061, ..., 0.0147, 0.0239, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0124, -0.0410, 0.0128, ..., -0.0254, -0.0033, -0.0173], + [ 0.0476, 0.0315, 0.0141, ..., -0.0148, -0.0069, 0.0049], + [-0.0096, -0.0270, 0.0089, ..., 0.0332, 0.0148, -0.0081], + ..., + [-0.0051, -0.0019, 0.0162, ..., -0.0177, 0.0171, -0.0132], + [ 0.0052, -0.0187, -0.0068, ..., 0.0248, -0.0142, -0.0076], + [-0.0143, -0.0047, 0.0150, ..., -0.0028, -0.0015, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0041, 0.0099, ..., 0.0255, -0.0253, -0.0129], + [-0.0042, -0.0139, -0.0160, ..., -0.0108, -0.0403, -0.0408], + [ 0.0175, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0113, 0.0052, 0.0322, ..., -0.0183, -0.0053, 0.0024], + [ 0.0053, -0.0057, 0.0162, ..., 0.0043, 0.0255, 0.0043], + [ 0.0270, 0.0085, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0075, ..., -0.0532, 0.0104, 0.0029], + [-0.0240, -0.0215, -0.0106, ..., -0.0050, 0.0097, -0.0168], + [ 0.0077, -0.0034, -0.0061, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0071, 0.0172, -0.0106, ..., -0.0300, 0.0005, 0.0435], + [ 0.0052, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0066], + [-0.0167, 0.0125, 0.0102, ..., -0.0247, 0.0120, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0046, -0.0081, 0.0317, ..., 0.0200, -0.0107, -0.0376], + [ 0.0356, 0.0378, 0.0135, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., 0.0027, 0.0050, -0.0322], + ..., + [ 0.0069, 0.0109, -0.0027, ..., 0.0366, -0.0206, 0.0374], + [-0.0054, -0.0080, 0.0038, ..., -0.0096, 0.0086, 0.0192], + [ 0.0099, -0.0178, -0.0038, ..., -0.0166, -0.0042, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0063, -0.0096, -0.0344, ..., -0.0049, 0.0347, -0.0181], + [ 0.0262, -0.0366, -0.0093, ..., 0.0216, 0.0349, -0.0067], + [-0.0100, -0.0286, 0.0056, ..., 0.0061, -0.0282, -0.0006], + ..., + [-0.0128, -0.0016, 0.0043, ..., -0.0089, 0.0165, 0.0076], + [-0.0322, -0.0082, -0.0170, ..., 0.0195, 0.0212, -0.0135], + [-0.0139, 0.0070, -0.0170, ..., -0.0129, 0.0059, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0043, -0.0065, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0115, 0.0374, ..., -0.0153, -0.0188, -0.0138], + ..., + [ 0.0262, 0.0003, -0.0069, ..., 0.0003, -0.0006, -0.0165], + [-0.0222, -0.0156, -0.0437, ..., -0.0065, -0.0165, -0.0282], + [-0.0130, -0.0146, 0.0029, ..., 0.0023, -0.0234, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0144, -0.0649, 0.0102], + [ 0.0249, 0.0205, -0.0050, ..., -0.0064, -0.0091, -0.0103], + [-0.0220, 0.0038, 0.0051, ..., 0.0306, 0.0381, -0.0036], + ..., + [ 0.0200, -0.0042, 0.0170, ..., 0.0118, 0.0320, -0.0157], + [ 0.0349, 0.0121, 0.0197, ..., 0.0398, 0.0315, -0.0217], + [-0.0081, 0.0256, 0.0078, ..., 0.0205, -0.0070, -0.0100]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0065, 0.0199, ..., -0.0063, 0.0276, -0.0312], + [ 0.0178, -0.0055, 0.0096, ..., 0.0113, -0.0398, 0.0293], + [ 0.0430, -0.0120, 0.0085, ..., -0.0029, -0.0067, 0.0060], + ..., + [ 0.0193, -0.0109, -0.0112, ..., -0.0049, 0.0330, 0.0143], + [-0.0122, -0.0542, -0.0205, ..., 0.0034, -0.0159, 0.0320], + [ 0.0303, -0.0039, -0.0425, ..., -0.0064, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0067, -0.0315, ..., 0.0525, -0.0052, -0.0349], + [-0.0031, -0.0032, -0.0156, ..., 0.0055, -0.0240, 0.0305], + [-0.0256, 0.0006, 0.0060, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0053, 0.0231, 0.0209, ..., 0.0073, 0.0050, 0.0461], + [-0.0078, 0.0006, 0.0060, ..., -0.0217, 0.0347, -0.0021], + [ 0.0051, 0.0234, -0.0025, ..., 0.0261, -0.0041, 0.0103]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0037, 0.0006], + [-0.0183, -0.0211, -0.0049, ..., -0.0145, 0.0146, -0.0162], + [-0.0060, -0.0160, 0.0062, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0098, 0.0064, ..., 0.0171, -0.0042, -0.0064], + [-0.0142, 0.0260, -0.0083, ..., 0.0067, 0.0104, -0.0166], + [ 0.0095, -0.0142, -0.0040, ..., 0.0062, 0.0078, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0074, 0.0143, -0.0065, ..., -0.0278, 0.0240, -0.0118], + [ 0.0012, 0.0188, -0.0160, ..., -0.0317, -0.0069, 0.0182], + [ 0.0071, 0.0310, -0.0093, ..., -0.0226, 0.0023, 0.0334], + ..., + [-0.0092, 0.0186, -0.0053, ..., 0.0228, 0.0063, 0.0050], + [ 0.0271, 0.0123, 0.0354, ..., -0.0062, 0.0198, 0.0143], + [ 0.0282, 0.0171, 0.0317, ..., -0.0113, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0110, -0.0265, -0.0115, ..., 0.0049, 0.0038, 0.0132], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0085], + [-0.0044, -0.0190, -0.0063, ..., 0.0166, -0.0135, 0.0137], + ..., + [-0.0200, -0.0009, 0.0057, ..., -0.0339, 0.0361, 0.0006], + [-0.0082, -0.0250, 0.0151, ..., -0.0035, -0.0162, 0.0220], + [-0.0075, 0.0022, 0.0205, ..., -0.0065, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0066, 0.0240, -0.0217, ..., 0.0172, 0.0045, 0.0109], + [-0.0386, 0.0103, 0.0045, ..., -0.0055, 0.0322, 0.0087], + [-0.0427, -0.0052, -0.0042, ..., 0.0062, 0.0110, -0.0135], + ..., + [-0.0085, -0.0044, -0.0264, ..., -0.0586, -0.0630, -0.0129], + [ 0.0042, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0106, -0.0610, 0.0162, ..., -0.0291, -0.0166, -0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0122, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0114], + [ 0.0112, -0.0101, -0.0327, ..., 0.0447, 0.0074, 0.0076], + [ 0.0053, -0.0175, 0.0070, ..., -0.0082, 0.0004, -0.0060], + ..., + [ 0.0295, -0.0042, -0.0103, ..., 0.0100, -0.0160, -0.0288], + [ 0.0085, 0.0253, -0.0177, ..., 0.0098, 0.0081, 0.0056], + [-0.0139, -0.0064, 0.0136, ..., -0.0128, 0.0277, 0.0086]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0061, -0.0276, -0.0271], + [ 0.0135, 0.0064, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0085, 0.0234, 0.0076, ..., 0.0464, 0.0066, 0.0309], + ..., + [-0.0123, -0.0175, 0.0327, ..., 0.0132, 0.0025, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, -0.0034, -0.0444, ..., 0.0058, -0.0283, 0.0029]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0067, -0.0104, 0.0148, ..., -0.0742, -0.0073, 0.0161], + [-0.0079, -0.0133, -0.0060, ..., 0.0056, 0.0049, 0.0239], + [ 0.0189, 0.0221, -0.0090, ..., 0.0049, 0.0408, 0.0049], + ..., + [ 0.0057, 0.0247, 0.0396, ..., 0.0057, -0.0304, -0.0183], + [ 0.0299, -0.0115, -0.0417, ..., -0.0049, -0.0376, 0.0041], + [-0.0210, 0.0168, -0.0003, ..., 0.0063, 0.0178, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[-0.0018, 0.0242, -0.0083, ..., -0.0245, 0.0064, -0.0058], + [ 0.0190, -0.0127, 0.0332, ..., -0.0146, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0079, ..., 0.0003, -0.0093, 0.0182], + ..., + [-0.0430, -0.0403, -0.0087, ..., 0.0135, -0.0141, -0.0055], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0095, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0061, 0.0122, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0122, ..., -0.0160, -0.0272, 0.0299], + [ 0.0095, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0076, ..., 0.0101, -0.0223, -0.0067], + ..., + [ 0.0192, 0.0292, -0.0026, ..., 0.0176, -0.0197, 0.0059], + [ 0.0265, 0.0063, 0.0067, ..., 0.0286, -0.0018, -0.0267], + [-0.0184, -0.0088, -0.0049, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0083, -0.0159, 0.0084, ..., 0.0198, 0.0059, 0.0277], + [ 0.0172, 0.0039, 0.0039, ..., -0.0035, 0.0021, 0.0022], + [-0.0059, 0.0184, 0.0111, ..., 0.0099, -0.0046, -0.0077], + ..., + [-0.0050, -0.0112, -0.0041, ..., -0.0096, -0.0050, 0.0118], + [-0.0427, -0.0391, -0.0192, ..., 0.0089, -0.0238, -0.0084], + [-0.0042, -0.0106, -0.0222, ..., -0.0164, 0.0342, 0.0090]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0087, -0.0161, 0.0119], + [ 0.0041, 0.0074, 0.0312, ..., 0.0074, 0.0286, 0.0317], + [ 0.0214, -0.0145, 0.0229, ..., -0.0082, 0.0056, 0.0277], + ..., + [ 0.0170, -0.0082, 0.0471, ..., 0.0072, 0.0076, -0.0046], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0067, -0.0203], + [ 0.0079, 0.0083, -0.0037, ..., -0.0089, -0.0074, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[ 0.0036, -0.0154, 0.0284, ..., -0.0104, 0.0135, -0.0042], + [ 0.0153, 0.0332, 0.0337, ..., 0.0223, -0.0165, 0.0051], + [-0.0068, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0069, 0.0077, 0.0280, ..., -0.0041, -0.0078, 0.0140], + [-0.0066, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0092, ..., 0.0064, 0.0167, -0.0086]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0079, ..., -0.0238, -0.0099, 0.0118], + [-0.0250, -0.0162, 0.0029, ..., -0.0085, -0.0020, -0.0369], + [ 0.0019, 0.0141, 0.0074, ..., -0.0452, 0.0014, 0.0119], + ..., + [ 0.0149, -0.0140, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0096, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0143, 0.0039, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[-0.0063, -0.0327, -0.0190, ..., 0.0078, 0.0013, -0.0209], + [ 0.0713, 0.0153, -0.0053, ..., -0.0035, 0.0186, 0.0002], + [-0.0227, -0.0094, 0.0188, ..., -0.0128, 0.0193, 0.0068], + ..., + [ 0.0114, 0.0464, 0.0082, ..., 0.0248, 0.0168, -0.0040], + [ 0.0381, -0.0038, 0.0063, ..., -0.0043, -0.0113, -0.0037], + [ 0.0292, 0.0044, 0.0209, ..., -0.0064, 0.0234, 0.0123]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0049, 0.0212, ..., 0.0305, 0.0136, 0.0105], + [-0.0077, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0049], + ..., + [ 0.0084, -0.0275, 0.0225, ..., 0.0055, -0.0189, -0.0105], + [-0.0125, 0.0047, -0.0058, ..., 0.0038, -0.0036, 0.0211], + [-0.0061, 0.0038, -0.0060, ..., -0.0110, 0.0264, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[ 4.6387e-03, -7.4768e-03, -1.0864e-02, ..., 1.8555e-02, + -2.1606e-02, 5.0964e-03], + [-3.8300e-03, 1.2207e-02, -2.3071e-02, ..., 1.2329e-02, + -1.1230e-02, -1.8311e-02], + [ 1.0681e-02, 1.0315e-02, 3.9041e-06, ..., 4.2480e-02, + 3.0884e-02, -2.7344e-02], + ..., + [ 3.3203e-02, 1.2390e-02, 1.4587e-02, ..., 5.3955e-02, + 4.1504e-02, 9.7046e-03], + [ 3.8605e-03, 8.5449e-03, -1.4038e-02, ..., 4.4250e-03, + -1.8921e-02, -2.3682e-02], + [-4.3030e-03, 3.5400e-02, 1.7700e-02, ..., 3.6133e-02, + -3.1128e-02, 6.9885e-03]], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0061, 0.0359, ..., -0.0090, -0.0128, -0.0022], + [ 0.0190, -0.0052, 0.0298, ..., 0.0125, -0.0043, -0.0200], + [ 0.0116, -0.0109, 0.0177, ..., -0.0248, 0.0229, 0.0014], + ..., + [-0.0082, 0.0022, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, -0.0047, -0.0291, ..., 0.0206, 0.0304, -0.0090], + [-0.0072, 0.0245, 0.0047, ..., -0.0413, 0.0216, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0062, -0.0157, -0.0065, ..., -0.0067, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0069], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, -0.0038], + ..., + [ 0.0046, -0.0049, 0.0258, ..., -0.0310, 0.0090, -0.0361], + [ 0.0061, 0.0308, 0.0542, ..., -0.0542, -0.0155, 0.0259], + [-0.0022, 0.0157, -0.0074, ..., -0.0018, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0079, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0107, ..., 0.0103, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0042, ..., 0.0123, -0.0043, 0.0089], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [ 0.0033, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0126, 0.0206, ..., 0.0043, -0.0122, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0096, -0.0231, ..., -0.0214, 0.0447, 0.0136], + [-0.0171, -0.0064, 0.0077, ..., 0.0043, 0.0114, 0.0598], + [-0.0047, 0.0267, -0.0033, ..., 0.0284, -0.0074, -0.0364], + ..., + [-0.0050, -0.0033, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0037, -0.0091, ..., 0.0211, 0.0138, -0.0076], + [ 0.0100, 0.0236, -0.0206, ..., -0.0134, 0.0046, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0064, 0.0294, ..., 0.0033, 0.0270, 0.0055], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0061, -0.0168], + [ 0.0201, -0.0121, 0.0124, ..., -0.0381, -0.0247, -0.0070], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0075], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0066, -0.0466], + [ 0.0327, 0.0284, -0.0048, ..., 0.0145, 0.0140, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0138, -0.0038, -0.0167, ..., -0.0034, 0.0170, 0.0108], + [ 0.0037, -0.0002, 0.0295, ..., -0.0172, 0.0009, 0.0135], + [ 0.0193, 0.0236, -0.0123, ..., 0.0049, -0.0369, 0.0239], + ..., + [ 0.0153, -0.0413, -0.0125, ..., -0.0354, 0.0059, -0.0320], + [-0.0266, -0.0062, -0.0327, ..., -0.0020, 0.0045, 0.0047], + [ 0.0298, 0.0356, -0.0192, ..., 0.0042, -0.0004, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0077, 0.0059, 0.0317, ..., 0.0049, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0079, -0.0096], + [-0.0166, 0.0073, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0050, 0.0260, ..., 0.0159, 0.0034, -0.0474], + [ 0.0038, 0.0070, -0.0074, ..., 0.0140, 0.0557, 0.0044], + [-0.0203, 0.0273, -0.0075, ..., -0.0330, -0.0016, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0072, ..., 0.0039, -0.0093, -0.0277], + [ 0.0057, -0.0061, -0.0058, ..., -0.0060, 0.0140, -0.0161], + [-0.0047, 0.0168, -0.0055, ..., -0.0092, 0.0124, 0.0036], + ..., + [ 0.0398, 0.0452, 0.0251, ..., -0.0067, -0.0330, -0.0012], + [-0.0620, -0.0040, 0.0371, ..., 0.0410, 0.0041, -0.0173], + [ 0.0126, -0.0039, -0.0270, ..., -0.0043, 0.0388, -0.0103]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0127, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0005, 0.0245], + [-0.0449, 0.0017, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0042, ..., 0.0041, 0.0288, 0.0127], + [ 0.0330, 0.0028, 0.0062, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0078, -0.0033, ..., -0.0060, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0066, 0.0040, 0.0527], + [ 0.0037, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0099, -0.0095, 0.0327], + [ 0.0135, 0.0113, -0.0067, ..., 0.0093, -0.0178, -0.0254], + [-0.0530, -0.0125, 0.0281, ..., -0.0242, -0.0128, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, 0.0029], + [-0.0052, -0.0025, -0.0286, ..., 0.0053, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0052, ..., 0.0018, 0.0083, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0099, 0.0150], + [ 0.0014, 0.0037, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0079, 0.0076, ..., 0.0092, -0.0111, 0.0041]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0075, -0.0038, ..., -0.0220, -0.0145, -0.0047], + [-0.0166, 0.0189, -0.0050, ..., 0.0096, 0.0016, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0043, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0070, ..., -0.0049, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0065, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0125, 0.0289, 0.0125, ..., 0.0135, 0.0065, -0.0105], + [ 0.0188, -0.0197, -0.0054, ..., 0.0284, -0.0342, -0.0066], + [-0.0525, -0.0214, 0.0153, ..., -0.0270, 0.0083, 0.0254], + ..., + [-0.0055, 0.0085, 0.0281, ..., 0.0129, -0.0035, -0.0065], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0046, 0.0033], + [-0.0273, -0.0747, 0.0208, ..., -0.0047, 0.0046, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0056, 0.0148, 0.0334, ..., 0.0205, -0.0273, -0.0020], + [-0.0165, 0.0110, -0.0135, ..., 0.0106, -0.0067, 0.0219], + [ 0.0046, 0.0227, -0.0055, ..., 0.0126, -0.0061, 0.0151], + ..., + [-0.0347, -0.0148, 0.0062, ..., 0.0040, 0.0030, -0.0093], + [-0.0073, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0039, -0.0500, -0.0232, ..., 0.0320, 0.0037, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0052, -0.0364], + [ 0.0220, -0.0026, -0.0115, ..., -0.0043, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0117, ..., 0.0444, 0.0049, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0067, 0.0192, -0.0393], + [-0.0102, -0.0244, 0.0275, ..., 0.0038, 0.0043, 0.0081], + [-0.0170, -0.0427, 0.0132, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0050, -0.0378, -0.0359, ..., 0.0552, -0.0075, -0.0160], + [ 0.0047, 0.0142, 0.0134, ..., 0.0055, -0.0153, 0.0354], + [ 0.0094, 0.0040, 0.0057, ..., 0.0028, -0.0249, -0.0136], + ..., + [ 0.0068, -0.0125, 0.0049, ..., -0.0095, -0.0079, -0.0036], + [-0.0073, 0.0439, 0.0166, ..., 0.0248, 0.0134, -0.0187], + [ 0.0175, 0.0061, 0.0049, ..., 0.0410, -0.0096, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0107, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0142, 0.0038, ..., -0.0035, -0.0244, 0.0356], + [ 0.0071, 0.0165, -0.0200, ..., 0.0033, 0.0237, -0.0078], + ..., + [-0.0197, 0.0054, 0.0386, ..., 0.0181, 0.0109, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, -0.0047, -0.0374], + [-0.0085, -0.0095, 0.0172, ..., 0.0243, -0.0219, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-3.4668e-02, 6.2866e-03, 1.9287e-02, ..., -2.2095e-02, + -1.4771e-02, -2.1118e-02], + [ 1.3062e-02, 1.1658e-02, 2.5269e-02, ..., 2.0386e-02, + -2.3193e-02, -3.2227e-02], + [ 2.5513e-02, -4.8523e-03, 4.1809e-03, ..., -5.7068e-03, + 5.4016e-03, 3.3691e-02], + ..., + [ 4.2236e-02, 2.4414e-03, 3.5156e-02, ..., -3.0396e-02, + -4.7207e-05, -2.0386e-02], + [ 3.0151e-02, 1.5503e-02, 9.9487e-03, ..., -7.2327e-03, + -2.5391e-02, -2.4536e-02], + [-6.4087e-04, -6.2561e-03, -2.0599e-03, ..., 4.1504e-03, + -1.9653e-02, -2.8198e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0042, 0.0266, ..., 0.0155, 0.0187, 0.0047], + [-0.0118, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0036, ..., -0.0172, -0.0092, 0.0231], + ..., + [-0.0295, -0.0136, -0.0036, ..., 0.0042, -0.0048, 0.0101], + [-0.0048, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0078], + [-0.0015, -0.0043, -0.0259, ..., -0.0030, -0.0141, -0.0115]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0040, 0.0064, -0.0140, ..., 0.0242, -0.0040, -0.0386], + [-0.0142, 0.0068, -0.0101, ..., -0.0369, -0.0151, 0.0157], + [-0.0267, -0.0045, -0.0074, ..., -0.0126, 0.0182, -0.0115], + ..., + [ 0.0172, -0.0038, 0.0144, ..., 0.0115, -0.0024, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., -0.0023, -0.0033, -0.0137], + [-0.0219, -0.0396, 0.0039, ..., -0.0054, -0.0037, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-3.3417e-03, 1.6846e-02, 1.9409e-02, ..., 3.4912e-02, + -4.1992e-02, 2.0142e-03], + [ 4.3945e-03, 2.7344e-02, 1.4526e-02, ..., -1.8433e-02, + -3.5248e-03, 1.5442e-02], + [ 1.4465e-02, -3.7231e-03, -7.3242e-03, ..., 2.4536e-02, + 3.5889e-02, 5.8289e-03], + ..., + [ 4.3213e-02, 1.4526e-02, -3.1982e-02, ..., 8.8215e-05, + 4.2114e-03, 2.6611e-02], + [ 2.2949e-02, 1.2573e-02, -1.9897e-02, ..., -7.5989e-03, + -3.1250e-02, -4.5654e-02], + [ 4.9133e-03, 1.0910e-03, -6.8359e-03, ..., 4.9133e-03, + -1.0437e-02, -1.5503e-02]], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0115, 0.0388, 0.0359], + [-0.0042, 0.0121, -0.0019, ..., 0.0123, -0.0046, 0.0068], + [-0.0277, 0.0011, -0.0181, ..., -0.0003, -0.0036, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0035, 0.0513, -0.0109], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0110], + [-0.0282, -0.0046, 0.0119, ..., -0.0055, -0.0038, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0033, -0.0128, ..., 0.0140, 0.0308, 0.0059], + [ 0.0120, -0.0332, -0.0320, ..., -0.0075, -0.0140, -0.0032], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, 0.0029, -0.0311, ..., 0.0124, 0.0059, 0.0081], + [ 0.0040, 0.0198, 0.0141, ..., -0.0222, -0.0143, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, 0.0049, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0090, -0.0035, -0.0027, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0049, ..., 0.0302, -0.0018, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0023, -0.0036]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0028, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0037, -0.0222, 0.0262], + ..., + [-0.0147, 0.0043, 0.0260, ..., -0.0042, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0066, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0035, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0112], + [-0.0278, 0.0041, -0.0374, ..., 0.0276, -0.0003, 0.0151], + [ 0.0089, 0.0237, -0.0039, ..., 0.0527, -0.0031, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0072, -0.0095], + [-0.0067, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0036, -0.0128, ..., 0.0063, -0.0371, -0.0064], + [ 0.0064, -0.0130, -0.0028, ..., -0.0204, -0.0013, 0.0057], + [-0.0112, -0.0195, -0.0053, ..., -0.0542, -0.0094, -0.0059], + ..., + [-0.0205, -0.0082, -0.0037, ..., -0.0199, 0.0044, 0.0058], + [-0.0220, -0.0302, -0.0105, ..., 0.0114, 0.0369, 0.0052], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0134, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[ 0.0017, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0034, 0.0056, 0.0238, ..., -0.0140, -0.0320, 0.0041], + [-0.0193, 0.0364, -0.0035, ..., -0.0076, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0065, -0.0232, -0.0464], + [-0.0107, -0.0298, -0.0195, ..., 0.0109, 0.0312, 0.0151], + [-0.0074, -0.0116, -0.0195, ..., 0.0034, 0.0107, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0149, ..., -0.0371, 0.0400, -0.0165], + [ 0.0047, -0.0239, -0.0123, ..., -0.0062, 0.0243, -0.0396], + [-0.0101, 0.0155, -0.0038, ..., -0.0277, 0.0042, 0.0037], + ..., + [ 0.0112, -0.0047, 0.0515, ..., 0.0327, -0.0120, -0.0077], + [-0.0030, 0.0170, 0.0063, ..., -0.0014, 0.0280, -0.0210], + [-0.0016, 0.0076, 0.0055, ..., 0.0153, -0.0044, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0063, 0.0081, -0.0077, ..., -0.0588, 0.0187, -0.0068], + [-0.0102, 0.0199, -0.0077, ..., 0.0080, -0.0332, -0.0026], + ..., + [-0.0051, -0.0464, 0.0277, ..., -0.0143, 0.0198, 0.0071], + [ 0.0162, -0.0053, 0.0082, ..., -0.0105, -0.0038, 0.0297], + [-0.0269, 0.0129, 0.0164, ..., 0.0086, 0.0176, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0033, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, 0.0028, ..., 0.0142, 0.0128, 0.0439], + [-0.0352, 0.0042, -0.0201, ..., 0.0330, 0.0309, -0.0049], + ..., + [-0.0033, 0.0123, 0.0153, ..., -0.0034, -0.0038, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0039, 0.0175], + [-0.0244, 0.0036, -0.0300, ..., 0.0051, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0046, -0.0019, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0029, ..., 0.0310, 0.0088, -0.0023], + [ 0.0033, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0109, -0.0237, ..., -0.0104, 0.0157, -0.0037], + [-0.0081, 0.0205, 0.0140, ..., 0.0262, -0.0195, 0.0077], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, 0.0040, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0018, 0.0034, -0.0093], + [-0.0009, 0.0019, 0.0398, ..., -0.0304, -0.0029, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0019], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0042, 0.0065, 0.0228, ..., -0.0322, 0.0315, -0.0012], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0050, 0.0291], + [ 0.0135, -0.0014, -0.0339, ..., 0.0074, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, -0.0024], + [-0.0166, -0.0043, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0065, -0.0039, ..., 0.0240, 0.0037, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, -0.0020], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0014, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0011, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0066, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0022], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, -0.0039, 0.0036], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, 0.0027, ..., -0.0164, -0.0020, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0033, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0027, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0056, -0.0231, -0.0255, ..., 0.0038, 0.0295, -0.0057], + [-0.0255, 0.0513, 0.0029, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0031, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0067, -0.0070, -0.0364, ..., -0.0175, 0.0306, -0.0078], + [-0.0051, -0.0195, -0.0012, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0075, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0072, -0.0193, -0.0140], + [-0.0166, 0.0062, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0133, -0.0056, 0.0145, ..., -0.0072, 0.0211, -0.0275], + [ 0.0078, -0.0057, -0.0234, ..., 0.0084, 0.0071, -0.0166], + [ 0.0225, -0.0251, 0.0052, ..., 0.0078, 0.0236, -0.0046], + ..., + [ 0.0186, 0.0143, -0.0046, ..., 0.0272, -0.0083, -0.0096], + [ 0.0077, 0.0035, 0.0072, ..., -0.0260, 0.0267, 0.0151], + [ 0.0103, -0.0076, 0.0212, ..., 0.0195, -0.0157, 0.0121]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0087, -0.0270, 0.0251, ..., 0.0096, 0.0049, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.3.proj.1.weight': tensor([[-0.0172, 0.0071, 0.0757, ..., -0.0610, 0.0327, -0.0320], + [-0.0278, 0.0029, -0.1128, ..., 0.0272, -0.0361, 0.0583], + [-0.0347, 0.0947, -0.0361, ..., -0.0166, 0.0172, -0.0332], + ..., + [ 0.0125, 0.0007, -0.0100, ..., 0.0219, -0.0376, -0.0036], + [-0.0239, -0.0574, 0.0586, ..., 0.0378, 0.0830, 0.0608], + [ 0.0176, -0.0205, 0.0076, ..., 0.0248, 0.0493, 0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_proj.weight': tensor([[-4.1199e-03, 5.4932e-03, 6.3782e-03, ..., 2.3346e-03, + -5.4359e-05, 8.1177e-03], + [-9.6893e-04, 8.4305e-04, -8.3923e-04, ..., -8.9111e-03, + -4.5586e-04, 2.2736e-03], + [ 4.7913e-03, -3.5553e-03, 3.2806e-03, ..., -8.0566e-03, + -4.3335e-03, 1.3809e-03], + ..., + [-2.3193e-03, -4.1504e-03, -7.0496e-03, ..., 1.2268e-02, + 4.0588e-03, -8.0566e-03], + [-9.6436e-03, -7.9346e-03, -1.2634e-02, ..., -1.3123e-02, + -5.9204e-03, -6.7139e-04], + [-1.7929e-03, -8.9111e-03, 6.1646e-03, ..., 9.9487e-03, + -4.2915e-04, 1.8005e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_norm.weight': tensor([ 4.7852e-02, -1.6968e-02, 4.7119e-02, 7.8613e-02, -1.3809e-03, + 4.0513e-08, -1.1362e-16, -1.8396e-21, -3.8598e-17, -1.1689e-19, + -5.3204e-21, 3.8257e-24, -6.4851e-22, -1.0727e-24, -2.7504e-23, + 3.6028e-25, -6.9384e-08, -4.4250e-03, -2.9395e-09, -8.8476e-09, + 7.6834e-09, 1.5924e-19, 2.1973e-02, 3.1281e-04, -5.5908e-02, + 4.2578e-01, -1.0889e-01, -3.9062e-03, 1.7822e-02, 8.7891e-02, + 1.0029e-18, 2.1537e-08, 2.6855e-03, -5.1737e-05, -5.5552e-05, + 5.1880e-04, -4.3074e-09, 8.7486e-14, -4.0118e-23, 2.0195e-26, + -7.4216e-27, 8.4703e-22, 3.2558e-21, -1.4343e-03, 1.9775e-02, + -3.8719e-04, -5.4688e-02, -5.0293e-02, 1.7969e-01, 2.9541e-02, + -5.5542e-03, 8.2397e-04, -3.9062e-03, 3.2596e-07, 7.0690e-17, + -7.6762e-22, 5.8981e-17, 1.3897e-22, -2.2334e-22, 9.1382e-27, + 2.7297e-22, -1.3944e-13, 1.9690e-27, 6.0269e-28, 1.2988e-01, + 1.4746e-01, 3.4570e-01, 7.7820e-03, 1.8311e-03, 2.5728e-08, + 7.2858e-17, 3.7481e-20, -1.4162e-18, -9.3512e-19, -7.4115e-20, + -1.3329e-25, 2.2658e-20, -5.5008e-23, 2.5312e-22, 7.6256e-25, + -1.3318e-07, -7.6771e-05, -2.6375e-06, 3.3424e-11, 2.0210e-07, + -1.2559e-15, -2.5635e-03, -1.6022e-03, 3.4570e-01, -1.5625e-02, + 4.1992e-01, 7.9102e-02, -8.3618e-03, 5.3711e-03, -1.2536e-18, + 1.3132e-07, 6.4087e-04, -9.7156e-06, -3.7193e-05, 5.3406e-05, + -7.4971e-08, 1.2546e-14, -2.7303e-25, -2.2877e-24, -4.2662e-27, + -1.5096e-23, 7.9939e-21, 1.2756e-02, 2.8076e-03, -1.2054e-03, + 3.9844e-01, 2.0703e-01, 1.5430e-01, -1.6113e-01, 7.7515e-03, + -4.9805e-02, -3.1128e-03, 1.5087e-07, 2.9707e-17, -5.7439e-21, + 3.4911e-17, 3.6528e-21, 7.3268e-20, 4.3823e-26, 1.4540e-24, + 1.9895e-11, 7.8065e-24, -1.0057e-25], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.self_attn.k_proj.weight': tensor([[-7.0496e-03, 4.8218e-03, -7.4768e-03, ..., -4.2114e-03, + -3.3264e-03, 2.1515e-03], + [-1.9287e-02, -1.4496e-03, 1.9226e-03, ..., 1.4114e-03, + -5.8289e-03, -1.9531e-03], + [-2.1219e-05, -1.2436e-03, 6.7139e-03, ..., 1.1826e-03, + 1.5488e-03, 8.5449e-03], + ..., + [ 2.5330e-03, 2.9602e-03, -2.6550e-03, ..., -5.2261e-04, + -1.1215e-03, 3.8910e-03], + [ 1.1368e-03, 1.0757e-03, -7.6599e-03, ..., -1.0529e-03, + -1.5354e-04, -1.1063e-03], + [ 1.4648e-03, 1.9684e-03, -3.1738e-03, ..., -5.8594e-03, + -2.7466e-04, 1.3962e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.k_norm.weight': tensor([ 9.4727e-02, 5.6641e-02, -5.5664e-02, 7.6660e-02, 1.5335e-03, + -2.1770e-08, -1.1189e-16, 1.7999e-20, -3.8381e-17, -1.6178e-19, + -8.7350e-21, -3.7740e-24, -2.6073e-21, 1.4605e-24, 1.8301e-23, + 6.4300e-25, -6.6590e-08, 4.1504e-03, 1.0477e-09, 8.8476e-09, + 7.5088e-09, -1.1181e-19, 1.6724e-02, 1.3504e-03, -1.5625e-02, + 4.2383e-01, -1.1816e-01, 7.5684e-02, 3.2349e-03, 4.5538e-05, + -1.4704e-18, 6.2864e-08, -2.6093e-03, -5.1498e-05, -5.5313e-05, + -5.4932e-04, -8.7894e-09, -7.1942e-14, 4.0325e-23, 2.8071e-26, + -6.6643e-27, -8.4372e-22, -5.1351e-21, 1.0193e-02, 1.8311e-02, + 2.2125e-04, 8.5449e-02, 1.2598e-01, 2.0801e-01, 1.6211e-01, + 7.1411e-03, 4.9072e-02, -3.7994e-03, -3.8557e-07, -4.6079e-18, + -1.6874e-21, 5.2313e-18, 4.4668e-22, -9.2644e-22, -1.1763e-26, + 2.6635e-22, -2.5757e-13, -2.0397e-26, -9.4663e-28, 8.1177e-03, + 1.4355e-01, 3.4766e-01, -4.5898e-02, 1.6556e-03, 4.7730e-08, + 5.6379e-17, 3.5999e-20, 1.6263e-18, 9.1480e-19, -7.3692e-20, + -6.8259e-26, 2.0964e-20, 5.5421e-23, -2.6304e-22, 7.8194e-25, + -1.6857e-07, -8.5354e-05, -2.6226e-06, 7.9012e-12, 1.9558e-07, + 1.2976e-15, -1.5564e-03, -3.9291e-04, 3.4570e-01, -1.6479e-02, + 4.1992e-01, 5.7983e-03, 1.7578e-02, 8.7402e-02, -1.7872e-19, + 6.9849e-08, 4.4632e-04, 1.0669e-05, -3.7432e-05, 4.8637e-05, + -7.9628e-08, -1.2379e-14, -1.7933e-25, -2.0550e-24, 4.2914e-27, + 1.5303e-23, -7.9409e-21, 8.2397e-03, 1.1292e-02, 7.4387e-04, + 4.0039e-01, 1.8359e-01, 1.4355e-01, 1.5625e-02, 7.4768e-03, + -1.1597e-03, -3.6621e-03, -8.1491e-08, -3.8164e-17, 5.7175e-21, + -3.7513e-17, -3.4940e-21, -7.1998e-20, 2.3426e-26, -2.2101e-24, + 1.9895e-11, 7.8065e-24, -1.0138e-25], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.self_attn.v_proj.weight': tensor([[-0.0040, 0.0033, -0.0051, ..., -0.0042, 0.0036, -0.0044], + [ 0.0073, -0.0018, 0.0019, ..., -0.0043, -0.0013, -0.0023], + [ 0.0152, 0.0034, -0.0048, ..., 0.0084, -0.0022, -0.0011], + ..., + [ 0.0050, 0.0029, -0.0007, ..., -0.0044, 0.0009, 0.0006], + [ 0.0009, 0.0044, -0.0021, ..., -0.0018, 0.0045, -0.0052], + [ 0.0008, -0.0037, 0.0006, ..., -0.0032, -0.0033, -0.0094]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.output_proj.weight': tensor([[ 0.0129, -0.0291, -0.0254, ..., 0.0038, 0.0062, 0.0021], + [ 0.0141, -0.0053, 0.0117, ..., 0.0043, 0.0077, 0.0028], + [-0.0046, 0.0078, 0.0023, ..., 0.0040, 0.0029, -0.0049], + ..., + [ 0.0094, -0.0002, -0.0016, ..., -0.0038, 0.0046, -0.0112], + [-0.0047, 0.0088, 0.0038, ..., -0.0010, 0.0015, 0.0010], + [-0.0056, 0.0040, -0.0090, ..., 0.0036, 0.0044, -0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_proj.weight': tensor([[-0.0025, -0.0159, 0.0013, ..., 0.0058, -0.0039, 0.0007], + [ 0.0072, 0.0210, 0.0029, ..., -0.0058, 0.0050, -0.0034], + [ 0.0085, 0.0194, 0.0049, ..., -0.0053, 0.0130, -0.0042], + ..., + [ 0.0072, 0.0052, 0.0026, ..., 0.0013, -0.0083, 0.0012], + [-0.0029, -0.0007, -0.0029, ..., -0.0027, 0.0109, -0.0044], + [ 0.0002, -0.0038, 0.0008, ..., -0.0022, 0.0015, 0.0057]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_norm.weight': tensor([ 2.0695e-04, 1.8677e-02, 2.3192e-11, 5.6458e-04, 3.6049e-04, + -9.7156e-06, -6.0499e-06, 2.2125e-03, 3.6377e-02, 2.6611e-02, + 1.4305e-04, 3.8330e-02, 5.5859e-01, 3.2196e-03, 8.4839e-03, + 1.1981e-05, 4.2969e-01, 6.8188e-05, 1.9360e-04, 1.1292e-02, + 3.8528e-04, 1.5640e-03, 9.1309e-02, 1.1902e-02, 4.0283e-03, + 6.3171e-03, 2.3682e-02, 3.0696e-06, 9.1553e-03, 2.3651e-04, + 3.6133e-01, 1.9165e-02, 6.0120e-03, 3.8281e-01, 5.3125e-01, + 9.4986e-04, 8.8882e-04, 2.5879e-02, 1.4258e-01, 8.6060e-03, + 1.6113e-02, 2.0142e-03, -4.4890e-07, 3.3875e-03, 1.8921e-03, + 4.7119e-02, 9.3994e-03, 4.0770e-05, 2.3041e-03, 2.0218e-04, + 1.9932e-04, 1.8552e-06, 1.4156e-06, -1.7762e-05, -3.5400e-03, + 1.3672e-02, 2.6367e-02, 1.3275e-03, 1.4954e-02, 1.4526e-02, + -7.0572e-05, -2.9057e-07, 2.9945e-04, 8.1253e-04, 1.1658e-02, + 4.3640e-03, 1.5640e-03, 2.2070e-01, 1.0803e-02, 2.4780e-02, + 3.9258e-01, 1.5259e-05, 4.6143e-02, 2.0020e-02, 3.7842e-03, + -1.9550e-04, 4.4556e-03, 6.7139e-03, 6.1035e-03, 9.2697e-04, + 9.8267e-03, 1.0071e-02, 1.6016e-01, 1.7929e-03, 2.3127e-05, + 1.1719e-02, -1.5945e-03, 1.1265e-05, 8.9111e-03, 4.7852e-01, + 5.0049e-03, 6.5231e-04, 1.4282e-02, 6.4697e-03, 2.5630e-06, + 4.1406e-01, 1.6928e-05, 1.7929e-04, -1.6964e-13, 6.5613e-03, + 3.2617e-01, -2.6524e-06, -4.1199e-03, 4.0283e-03, 3.4912e-02, + 6.1512e-05, -2.4199e-05, 1.8477e-05, 1.9775e-02, 1.0132e-02, + 1.3504e-03, -3.4332e-05, 9.3994e-03, 4.8218e-03, 2.2430e-03, + 7.1335e-04, 2.0874e-02, 1.7471e-03, -9.5367e-04, 5.3101e-03, + 4.9353e-05, 4.5471e-03, 9.1934e-04, 1.2054e-03, 1.5503e-02, + 1.4210e-04, 2.7161e-03, 1.8501e-04], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_proj.weight': tensor([[-0.0022, 0.0007, -0.0010, ..., -0.0041, 0.0014, -0.0015], + [-0.0013, -0.0021, -0.0042, ..., -0.0038, -0.0054, -0.0001], + [ 0.0043, -0.0015, -0.0052, ..., -0.0038, -0.0040, -0.0036], + ..., + [ 0.0038, -0.0025, 0.0025, ..., 0.0010, 0.0006, -0.0012], + [ 0.0008, 0.0005, -0.0023, ..., -0.0032, 0.0006, 0.0013], + [ 0.0017, 0.0019, -0.0031, ..., -0.0045, -0.0006, 0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_norm.weight': tensor([ 1.6499e-04, 1.8677e-02, -2.0009e-11, 2.2125e-04, 3.4523e-04, + 1.2815e-05, 6.0499e-06, 2.1667e-03, 3.5645e-02, 2.6611e-02, + 1.5354e-04, 3.8574e-02, 5.6250e-01, 3.1738e-03, 8.0566e-03, + 9.1195e-06, 4.2969e-01, 6.8188e-05, -1.1015e-04, 1.1230e-02, + 6.5994e-04, 1.5945e-03, 9.1309e-02, 1.2024e-02, 3.8757e-03, + 6.3171e-03, 2.3438e-02, 2.9057e-06, 8.5449e-03, 2.3651e-04, + 3.6133e-01, 1.8921e-02, 5.9509e-03, 3.8281e-01, 5.3125e-01, + 8.7738e-04, 1.0223e-03, 2.6123e-02, 1.4355e-01, 8.7280e-03, + 1.6113e-02, 2.0142e-03, -4.8429e-07, 3.1891e-03, 1.3733e-03, + 4.6875e-02, 9.3384e-03, -1.4901e-05, 2.3193e-03, -7.9155e-05, + 2.0218e-04, 1.2815e-06, 1.5944e-06, 2.4915e-05, 3.9673e-03, + 1.3550e-02, 2.6245e-02, 1.3504e-03, 1.5259e-02, 1.4587e-02, + -7.7248e-05, -5.5507e-07, -3.1471e-04, 7.4005e-04, 1.1719e-02, + 4.3335e-03, 1.5640e-03, 2.2070e-01, 1.0742e-02, 2.4902e-02, + 3.9258e-01, 1.5259e-05, 4.4434e-02, 2.0020e-02, 3.7994e-03, + 1.9455e-04, 3.6926e-03, 6.6833e-03, 6.1646e-03, 9.1171e-04, + 9.7656e-03, 1.0071e-02, 1.6016e-01, 1.5564e-03, -1.7524e-05, + 1.1719e-02, 1.9379e-03, 1.1861e-05, 8.9722e-03, 4.7852e-01, + 4.9744e-03, 6.1798e-04, 1.4099e-02, 6.4697e-03, 2.2501e-06, + 4.1406e-01, 4.2021e-06, 7.7724e-05, 2.2116e-13, 6.5613e-03, + 3.2812e-01, 2.6673e-06, 4.2114e-03, 4.0283e-03, 3.4912e-02, + 6.2943e-05, 3.9101e-05, 1.8835e-05, 1.9165e-02, 1.0193e-02, + 1.5945e-03, 3.5048e-05, 9.3384e-03, 4.7607e-03, 2.6093e-03, + -6.7139e-04, 2.0874e-02, 1.8539e-03, 1.2131e-03, 5.3101e-03, + 3.6955e-05, 4.6997e-03, 9.1171e-04, 1.1749e-03, 1.5076e-02, + -1.4114e-04, 2.8076e-03, -5.2452e-05], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.v_proj.weight': tensor([[-7.5989e-03, -1.1215e-03, 1.2329e-02, ..., -7.8964e-04, + 5.2490e-03, -1.0742e-02], + [-2.1729e-02, -2.2278e-03, 5.7068e-03, ..., -3.0060e-03, + -8.9111e-03, -2.3956e-03], + [-9.3994e-03, -1.8768e-03, 4.5776e-03, ..., -4.2114e-03, + 1.4771e-02, -1.2085e-02], + ..., + [ 1.0010e-02, -5.7220e-04, 4.4346e-05, ..., -8.2779e-04, + 3.7384e-03, 1.9897e-02], + [-7.6294e-03, 1.6708e-03, 1.0803e-02, ..., 1.3046e-03, + -1.7334e-02, -1.2360e-03], + [-1.6846e-02, -1.8082e-03, 8.1177e-03, ..., -1.2665e-03, + -7.4768e-03, -2.2583e-02]], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.output_proj.weight': tensor([[-0.0013, -0.0039, 0.0039, ..., 0.0077, 0.0042, 0.0098], + [ 0.0042, 0.0090, -0.0073, ..., 0.0137, 0.0020, -0.0074], + [ 0.0077, -0.0078, -0.0010, ..., 0.0038, -0.0059, -0.0006], + ..., + [-0.0054, -0.0056, -0.0043, ..., -0.0075, 0.0052, -0.0055], + [ 0.0081, 0.0005, 0.0091, ..., -0.0078, -0.0079, -0.0040], + [ 0.0175, -0.0094, -0.0039, ..., -0.0023, -0.0044, 0.0064]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_img.weight': tensor([[ 0.0078, 0.0121, 0.0018, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0074, 0.0249, 0.0291], + [-0.0095, 0.0080, -0.0121, ..., -0.0342, -0.0327, 0.0065], + ..., + [ 0.0051, -0.0067, 0.0038, ..., 0.0084, 0.0286, -0.0165], + [ 0.0078, -0.0067, -0.0101, ..., -0.0095, 0.0140, -0.0309], + [ 0.0242, -0.0101, 0.0255, ..., -0.0227, 0.0075, -0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, 0.0018], + [ 0.0041, 0.0120, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0080, ..., -0.0056, -0.0057, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, 0.0045, 0.0052], + [-0.0160, 0.0042, 0.0036, ..., 0.0133, -0.0173, 0.0024], + [ 0.0038, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0054, -0.0364, ..., -0.0225, -0.0145, -0.0052], + [-0.0157, -0.0052, -0.0364, ..., 0.0095, 0.0111, -0.0023], + [ 0.0037, 0.0615, 0.0018, ..., -0.0205, -0.0306, -0.0115], + ..., + [-0.0043, 0.0042, -0.0099, ..., -0.0211, -0.0044, 0.0288], + [ 0.0266, -0.0117, 0.0047, ..., -0.0420, -0.0282, -0.0014], + [ 0.0286, 0.0325, 0.0094, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.mlp.layer1.weight': tensor([[ 0.0036, -0.0023, -0.0013, ..., -0.0089, 0.0038, 0.0076], + [-0.0030, 0.0058, -0.0032, ..., -0.0057, 0.0044, 0.0010], + [-0.0087, 0.0012, -0.0061, ..., -0.0046, 0.0040, 0.0060], + ..., + [-0.0016, -0.0054, 0.0029, ..., -0.0075, -0.0067, 0.0027], + [-0.0028, 0.0018, -0.0085, ..., 0.0128, 0.0232, 0.0038], + [ 0.0034, 0.0015, -0.0031, ..., 0.0009, -0.0068, 0.0004]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.mlp.layer2.weight': tensor([[-2.4109e-03, 7.9346e-03, -9.7656e-03, ..., -1.1215e-03, + 2.0599e-03, 1.2268e-02], + [-1.8005e-03, 6.0425e-03, 3.6011e-03, ..., -1.6846e-02, + -2.3071e-02, 4.7607e-03], + [-7.0035e-06, -8.9722e-03, -4.5166e-03, ..., -5.1575e-03, + -7.6294e-03, -1.2573e-02], + ..., + [ 2.1515e-03, 1.1902e-03, -9.5215e-03, ..., 2.8381e-03, + 8.3008e-03, -7.6904e-03], + [-1.0437e-02, 5.0354e-03, 6.3324e-04, ..., 1.0742e-02, + 3.3447e-02, 5.6152e-03], + [-3.8300e-03, 1.1230e-02, 5.2490e-03, ..., 7.4768e-03, + 5.3406e-03, 2.8610e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0476, -0.0464, -0.0449, ..., 0.0027, 0.0028, 0.0029], + [ 0.0457, 0.0447, 0.0435, ..., -0.0021, -0.0022, -0.0023], + [ 0.0454, 0.0449, 0.0439, ..., -0.0023, -0.0024, -0.0024], + ..., + [ 0.0479, 0.0471, 0.0454, ..., -0.0021, -0.0022, -0.0022], + [ 0.0471, 0.0459, 0.0444, ..., -0.0020, -0.0020, -0.0019], + [ 0.0471, 0.0461, 0.0449, ..., -0.0027, -0.0029, -0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0021, 0.0021, 0.0022, ..., 0.0049, 0.0047, 0.0020], + [-0.0017, 0.0016, 0.0016, ..., 0.0049, 0.0052, 0.0017], + [-0.0059, 0.0059, 0.0054, ..., 0.0074, 0.0082, 0.0057], + ..., + [-0.0452, 0.0525, 0.0515, ..., 0.0515, 0.0488, 0.0452], + [ 0.0047, 0.0081, 0.0080, ..., -0.0068, -0.0065, -0.0046], + [ 0.0081, -0.0093, -0.0092, ..., -0.0067, -0.0059, -0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5847e-03, 5.9509e-03, 7.9346e-03, ..., -6.8665e-04, + -6.1417e-04, -7.8201e-04], + [-3.8086e-02, -3.6621e-02, -3.5889e-02, ..., 3.4332e-04, + 3.5286e-04, 3.6049e-04], + [ 4.2725e-02, 4.1016e-02, 4.0527e-02, ..., -1.5259e-05, + -5.3346e-06, -2.0266e-05], + ..., + [ 3.7842e-02, 3.7109e-02, 3.6377e-02, ..., -4.2152e-04, + -3.5667e-04, -3.9291e-04], + [ 4.9561e-02, 4.8096e-02, 4.7363e-02, ..., -7.9632e-05, + -4.7386e-06, -6.6757e-05], + [ 3.8574e-02, 3.7354e-02, 3.6865e-02, ..., 6.9141e-05, + 1.4782e-04, 9.6321e-05]], dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0070, -0.0059, 0.0048, ..., 0.0062, 0.0064, 0.0053], + [-0.0058, -0.0038, -0.0053, ..., 0.0038, 0.0035, 0.0025], + [-0.0043, -0.0044, -0.0085, ..., 0.0043, -0.0074, -0.0071], + ..., + [-0.0078, -0.0134, 0.0259, ..., 0.0132, 0.0273, 0.0216], + [ 0.0114, 0.0058, 0.0103, ..., -0.0056, 0.0052, 0.0107], + [-0.0127, 0.0037, -0.0052, ..., -0.0037, -0.0037, -0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_mlp.1.weight': tensor([[-0.0050, -0.0051, -0.0050, ..., 0.0005, 0.0007, 0.0007], + [ 0.0530, 0.0510, 0.0498, ..., -0.0020, -0.0024, -0.0024], + [ 0.0444, 0.0432, 0.0422, ..., -0.0033, -0.0036, -0.0036], + ..., + [ 0.0894, 0.0869, 0.0859, ..., -0.0005, -0.0007, -0.0007], + [ 0.0050, 0.0049, 0.0050, ..., -0.0005, -0.0006, -0.0005], + [ 0.0640, 0.0625, 0.0605, ..., -0.0010, -0.0010, -0.0010]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_mlp.2.weight': tensor([[-0.0010, 0.0074, 0.0080, ..., 0.0094, 0.0010, 0.0070], + [ 0.0007, 0.0026, 0.0031, ..., 0.0031, -0.0007, 0.0008], + [ 0.0004, 0.0029, 0.0034, ..., 0.0034, -0.0004, 0.0034], + ..., + [ 0.0072, 0.0603, 0.0820, ..., 0.0562, -0.0073, 0.0408], + [ 0.0059, 0.0070, 0.0063, ..., 0.0101, -0.0056, 0.0061], + [-0.0122, -0.0049, -0.0046, ..., -0.0051, 0.0120, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.before_proj.weight': tensor([[ 1.0431e-05, 1.1215e-03, 9.0027e-04, ..., -2.1820e-03, + 2.7657e-04, 3.4027e-03], + [ 3.8147e-03, 5.9509e-04, 1.0681e-02, ..., 1.2207e-02, + -7.5073e-03, -2.5635e-03], + [-1.1215e-03, -3.4637e-03, 2.3346e-03, ..., 1.5030e-03, + -9.9945e-04, -7.0190e-04], + ..., + [ 5.7373e-03, 2.3193e-03, -4.4060e-04, ..., -1.9684e-03, + -1.2207e-04, -5.2490e-03], + [ 2.5635e-03, -4.5166e-03, -4.9133e-03, ..., -1.8692e-03, + 4.3030e-03, -1.2512e-03], + [ 4.6387e-03, 3.4142e-04, -9.4604e-04, ..., 2.0905e-03, + -4.0588e-03, -5.6839e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.0.before_proj.bias': tensor([-0.0016, 0.0080, -0.0006, ..., -0.0053, 0.0027, 0.0032], + dtype=torch.bfloat16), 'net.control_blocks_3.0.after_proj.weight': tensor([[-0.0035, 0.0068, 0.0042, ..., -0.0038, 0.0090, 0.0054], + [-0.0079, -0.0034, 0.0029, ..., -0.0081, 0.0088, -0.0048], + [ 0.0004, -0.0017, -0.0006, ..., -0.0027, 0.0028, -0.0023], + ..., + [ 0.0025, -0.0022, 0.0061, ..., 0.0028, 0.0021, -0.0047], + [ 0.0118, -0.0138, -0.0078, ..., -0.0090, -0.0025, 0.0232], + [ 0.0028, 0.0028, 0.0025, ..., 0.0020, -0.0011, 0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.after_proj.bias': tensor([ 0.0344, -0.0278, 0.0008, ..., -0.0009, -0.0157, 0.0364], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_proj.weight': tensor([[-0.0245, -0.0016, -0.0092, ..., -0.0034, 0.0038, -0.0042], + [-0.0108, -0.0238, -0.0061, ..., -0.0143, -0.0064, -0.0309], + [ 0.0024, 0.0037, -0.0109, ..., 0.0135, 0.0092, -0.0177], + ..., + [ 0.0058, 0.0115, 0.0014, ..., 0.0066, -0.0047, -0.0063], + [ 0.0062, 0.0038, 0.0052, ..., 0.0110, 0.0022, 0.0024], + [-0.0030, -0.0042, 0.0009, ..., -0.0003, -0.0014, 0.0046]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_norm.weight': tensor([ 1.6211e-01, 5.1953e-01, 9.4727e-02, 6.3672e-01, 5.7031e-01, + 4.7265e-08, -2.4872e-03, -2.4011e-10, 2.0339e-13, -1.2306e-28, + 1.0193e-02, 3.3881e-19, -1.1181e-19, -2.1546e-33, 3.7748e-15, + -2.8407e-33, 1.8066e-01, -9.9182e-04, 4.8065e-04, 7.3547e-03, + 1.1921e-05, 5.1117e-04, 1.9727e-01, 7.4219e-01, 7.6172e-01, + 3.8867e-01, -5.8350e-02, 4.3359e-01, -1.3428e-02, 1.6602e-02, + -2.4128e-04, 1.0991e-14, -1.3123e-02, 6.1989e-06, 5.6458e-04, + -3.1433e-03, 3.5390e-08, 2.6090e-15, 4.0531e-05, 4.1567e-13, + 8.4128e-12, -3.3329e-29, 1.7975e-07, 8.5156e-01, 2.4707e-01, + 2.4121e-01, 2.6562e-01, 7.4609e-01, 1.8555e-01, -4.4189e-02, + 1.1902e-02, 1.8359e-01, 1.4305e-04, -1.0315e-02, -1.3924e-04, + 8.0945e-11, 1.6846e-02, -6.3109e-29, 2.7008e-03, -9.2699e-18, + 9.9487e-03, 1.9560e-34, -1.4572e-15, -1.5247e-26, 3.1445e-01, + 1.9922e-01, 4.9414e-01, 1.5527e-01, 9.0332e-02, 2.5495e-08, + 1.1719e-02, 8.4401e-10, 3.4106e-13, -9.5526e-31, -6.5308e-03, + 1.3485e-18, 1.7047e-20, -3.7074e-33, -2.8644e-14, -3.8519e-34, + -5.4932e-03, -2.9206e-05, -1.2131e-03, 2.9755e-03, 6.2943e-05, + 4.2725e-02, 8.1641e-01, 2.4316e-01, 2.2852e-01, 7.5000e-01, + 2.5586e-01, 2.5391e-01, -1.4893e-02, 4.1809e-03, -1.3306e-02, + 2.4203e-14, 1.0193e-02, 2.1338e-05, 3.3140e-05, 3.3417e-03, + 1.4043e-09, -5.7554e-13, 2.3651e-03, -8.5348e-16, 2.1316e-14, + 2.6770e-32, -1.9222e-06, 2.1582e-01, 7.8125e-01, 7.2656e-01, + 7.3438e-01, 2.1973e-01, 3.7305e-01, -5.6076e-04, 2.4414e-01, + 1.1353e-02, 2.4986e-04, -8.4839e-03, -1.3428e-03, 7.3123e-10, + 2.1606e-02, -5.9165e-29, 3.3875e-03, -1.1350e-19, -6.1035e-04, + 1.2488e-34, -9.4369e-15, -1.3545e-23], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.self_attn.k_proj.weight': tensor([[ 0.0018, -0.0083, 0.0069, ..., -0.0095, -0.0078, -0.0113], + [ 0.0317, -0.0110, 0.0075, ..., -0.0096, -0.0066, -0.0208], + [ 0.0042, 0.0084, 0.0037, ..., 0.0049, 0.0028, 0.0038], + ..., + [ 0.0018, -0.0026, -0.0050, ..., -0.0085, -0.0002, 0.0064], + [-0.0094, 0.0029, 0.0021, ..., 0.0157, 0.0093, -0.0103], + [ 0.0045, -0.0045, -0.0019, ..., -0.0126, -0.0001, 0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.k_norm.weight': tensor([ 2.1973e-01, 5.3906e-01, -8.6914e-02, 6.4844e-01, 5.7422e-01, + -3.9581e-08, -8.8501e-03, 4.4929e-10, -1.4477e-13, -5.2854e-29, + -1.1414e-02, -5.6243e-19, -1.0927e-19, -5.6273e-34, 3.0809e-15, + -9.8704e-34, 1.7969e-01, -9.8419e-04, -4.8065e-04, 4.8584e-02, + 1.1861e-05, 7.0095e-05, 2.0508e-01, 7.5000e-01, 7.6562e-01, + 1.4746e-01, 1.9629e-01, 4.9414e-01, -1.1292e-02, 1.6479e-02, + 8.1787e-03, 1.7097e-14, 1.5259e-02, 3.3230e-06, 5.8365e-04, + 3.7842e-02, 3.3993e-08, 5.4401e-15, 4.0293e-05, 4.2100e-13, + -8.4697e-12, -3.3921e-29, -6.0350e-07, 8.4766e-01, 2.3438e-01, + 2.1210e-03, 6.4941e-02, 7.4609e-01, 9.1797e-02, 3.6774e-03, + -2.4609e-01, 7.7637e-02, 1.4019e-04, -9.6436e-03, -1.0300e-04, + 1.2915e-10, 1.7578e-02, -6.1531e-29, -2.4567e-03, -7.9689e-18, + 9.7656e-03, 7.8993e-35, -4.8017e-15, 5.9373e-26, 3.0469e-01, + 2.3535e-01, 4.9805e-01, 1.2988e-01, -9.4238e-02, -3.0734e-08, + -7.0496e-03, 3.2742e-10, 2.4691e-13, 1.4298e-29, 1.6403e-03, + 1.3891e-18, 1.8423e-20, -6.9213e-34, -2.8200e-14, -8.5764e-35, + 4.5776e-03, -4.0293e-05, -1.2054e-03, 2.8076e-03, 6.2466e-05, + 4.2969e-02, 8.1250e-01, 1.2988e-01, 1.7969e-01, 7.6953e-01, + 2.2656e-01, 2.3047e-01, 1.6235e-02, -3.2501e-03, 9.4604e-03, + -2.2760e-14, 5.4626e-03, -2.2054e-05, -4.7386e-06, 3.1494e-02, + 1.3461e-09, -5.2935e-13, -2.4109e-03, -2.1927e-15, 9.3814e-15, + 5.4696e-32, -1.9222e-06, 2.5781e-01, 7.8516e-01, 7.3047e-01, + 7.4219e-01, -5.1270e-02, 4.0625e-01, -4.3701e-02, 5.5542e-03, + -1.5527e-01, 2.5558e-04, 9.0332e-03, -1.3428e-03, -7.2396e-10, + -1.6846e-02, -6.0348e-29, -3.0823e-03, -1.8465e-19, -5.7220e-04, + 9.7049e-35, -8.5487e-15, -1.1529e-23], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.self_attn.v_proj.weight': tensor([[ 5.7068e-03, 4.6387e-03, 6.9275e-03, ..., -1.5198e-02, + 3.6621e-03, 9.0599e-05], + [ 6.2561e-03, -3.5553e-03, 1.4343e-02, ..., -2.3193e-03, + -2.3956e-03, 3.4523e-04], + [ 4.8828e-03, 4.3030e-03, 3.4790e-03, ..., -6.3782e-03, + 5.6763e-03, -7.4463e-03], + ..., + [ 8.7891e-03, -1.0834e-03, 1.0437e-02, ..., 8.3618e-03, + 3.5248e-03, -1.6724e-02], + [-6.5308e-03, 4.5166e-03, 6.0425e-03, ..., 3.4180e-02, + -2.7924e-03, 5.6458e-03], + [-3.3875e-03, 1.0757e-03, -2.8076e-03, ..., 1.3245e-02, + -4.1199e-03, 4.6692e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.output_proj.weight': tensor([[-5.7373e-03, 7.4768e-03, 7.9155e-05, ..., 1.1292e-02, + -1.1719e-02, 7.5073e-03], + [ 1.1047e-02, -3.5553e-03, 1.7578e-02, ..., 4.7913e-03, + -1.2207e-02, 9.5215e-03], + [ 1.9531e-03, 2.9564e-04, 2.3499e-03, ..., 7.7820e-04, + 1.1841e-02, 4.6692e-03], + ..., + [ 1.1047e-02, 4.1504e-03, 1.1749e-03, ..., -4.1809e-03, + -7.9956e-03, -7.2937e-03], + [-7.1106e-03, 8.9722e-03, 4.9438e-03, ..., 4.0588e-03, + -3.7079e-03, 3.6926e-03], + [-1.1475e-02, 9.1553e-03, 1.1597e-02, ..., -2.4719e-03, + 8.1177e-03, -6.1340e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_proj.weight': tensor([[-0.0058, 0.0116, 0.0141, ..., -0.0040, 0.0054, 0.0082], + [ 0.0051, -0.0135, -0.0170, ..., 0.0056, -0.0195, -0.0311], + [ 0.0070, 0.0134, -0.0043, ..., -0.0032, 0.0034, -0.0074], + ..., + [ 0.0020, 0.0212, -0.0030, ..., 0.0063, 0.0024, -0.0049], + [-0.0104, 0.0107, 0.0090, ..., -0.0048, 0.0112, 0.0060], + [ 0.0151, -0.0081, -0.0079, ..., -0.0070, -0.0042, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_norm.weight': tensor([ 1.9775e-02, 4.4727e-01, 2.8320e-02, 4.8578e-06, 3.5048e-05, + 3.2507e-13, 2.3651e-03, 5.1172e-01, 2.6855e-03, 3.7695e-01, + 3.6133e-02, 1.2512e-02, 6.3171e-03, 1.2695e-02, 2.7466e-02, + 1.5381e-02, 1.2207e-02, 7.8201e-05, 2.8163e-06, 3.7109e-02, + 2.5391e-01, 2.5879e-02, 2.4915e-05, 2.4414e-02, 9.4604e-03, + 5.1737e-05, 4.1962e-05, 4.5166e-02, 1.3542e-04, 3.8281e-01, + -4.8399e-05, -1.2684e-04, 5.1880e-03, 8.8215e-05, 2.7008e-03, + 3.2806e-03, 1.9651e-07, 4.1211e-01, 1.0315e-02, 5.4932e-03, + 6.9336e-02, 1.2268e-02, 1.6846e-02, 9.5844e-05, 1.1826e-03, + 1.4526e-02, 3.8910e-03, 2.2217e-02, 2.2461e-02, 1.9775e-02, + 1.1414e-02, 3.8281e-01, 1.2360e-03, 1.9653e-02, 2.6489e-02, + 3.3875e-03, 1.0205e-01, 1.0132e-02, 2.3926e-02, 1.2573e-02, + 1.6975e-04, 5.3101e-03, 1.2878e-02, -1.7738e-04, 5.6152e-03, + 5.5542e-03, -1.0133e-05, 3.3188e-04, 1.6235e-02, 7.9590e-02, + 6.0425e-03, 8.1177e-03, 6.0547e-02, 4.1260e-02, 6.9275e-03, + 5.2185e-03, 7.6771e-05, 2.5879e-02, 3.2227e-02, 2.6489e-02, + 3.9673e-03, 2.0508e-02, 3.1738e-02, -4.5868e-08, 1.9653e-02, + 1.3733e-02, 2.7895e-05, 4.5410e-02, 1.1475e-02, 3.4943e-03, + 1.3916e-02, 7.1335e-04, 5.0049e-03, 2.3193e-02, 1.2878e-02, + -1.3256e-04, 1.3245e-02, 4.0283e-02, 2.0264e-02, 4.8096e-02, + 3.9551e-02, 3.1128e-02, 1.7944e-02, 3.3140e-05, -4.7493e-04, + 3.2616e-04, 1.0864e-02, 2.0264e-02, 1.8799e-02, 1.9287e-02, + 3.7384e-03, 1.3977e-02, 3.6865e-02, 4.9414e-01, -1.6391e-06, + 2.5269e-02, 6.9580e-03, 1.5991e-02, 1.4572e-03, 1.1182e-01, + 5.0068e-06, 2.2411e-05, 3.4571e-05, 1.1475e-02, 1.5015e-02, + 1.5625e-02, 2.3804e-02, 3.1494e-02], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_proj.weight': tensor([[-0.0017, -0.0033, 0.0020, ..., -0.0039, 0.0030, 0.0041], + [ 0.0123, -0.0035, 0.0190, ..., -0.0018, 0.0317, 0.0050], + [ 0.0035, 0.0036, 0.0027, ..., 0.0023, 0.0099, -0.0018], + ..., + [-0.0064, -0.0049, -0.0075, ..., -0.0028, 0.0013, 0.0076], + [-0.0078, -0.0053, -0.0181, ..., -0.0019, 0.0045, 0.0051], + [-0.0024, -0.0004, -0.0070, ..., 0.0022, 0.0089, -0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_norm.weight': tensor([ 2.0142e-02, 4.4727e-01, 2.8198e-02, 4.8876e-06, 3.5048e-05, + -3.1086e-13, 2.3651e-03, 5.1172e-01, 2.7618e-03, 3.7695e-01, + 3.7109e-02, 1.2512e-02, 6.2866e-03, 1.2817e-02, 2.7588e-02, + 1.5259e-02, 1.2451e-02, 7.7248e-05, 3.7402e-06, 3.7354e-02, + 2.5391e-01, 2.5146e-02, 2.4319e-05, 2.4658e-02, 9.5215e-03, + 5.0783e-05, 4.4346e-05, 4.4678e-02, 1.3733e-04, 3.8477e-01, + -4.8637e-05, -1.2684e-04, 5.0659e-03, 8.7738e-05, 2.7313e-03, + 3.3112e-03, 2.0582e-07, 4.1211e-01, 1.0315e-02, 5.4626e-03, + 6.8359e-02, 1.2207e-02, 1.6724e-02, 9.5367e-05, 1.1063e-03, + 1.4526e-02, 3.9368e-03, 2.1851e-02, 2.2461e-02, 2.0020e-02, + 1.1475e-02, 3.8281e-01, 1.1444e-03, 1.9287e-02, 2.6123e-02, + 3.3112e-03, 1.0205e-01, 1.0132e-02, 2.3438e-02, 1.2512e-02, + 1.6785e-04, 5.3406e-03, 1.3000e-02, -1.7738e-04, 5.6152e-03, + 5.4626e-03, -8.5235e-06, 3.1662e-04, 1.6357e-02, 7.9590e-02, + 6.0120e-03, 8.1177e-03, 6.0059e-02, 4.1260e-02, 6.9885e-03, + 5.3406e-03, 7.7248e-05, 2.5513e-02, 3.1738e-02, 2.6489e-02, + 3.9673e-03, 2.0386e-02, 3.1250e-02, -4.4005e-08, 1.9531e-02, + 1.3672e-02, 2.8133e-05, 4.5410e-02, 1.1902e-02, 3.6774e-03, + 1.4282e-02, 7.1716e-04, 5.0049e-03, 2.3071e-02, 1.2756e-02, + -1.3447e-04, 1.2695e-02, 3.9795e-02, 2.0264e-02, 4.8096e-02, + 4.0283e-02, 3.1128e-02, 1.8188e-02, 3.4094e-05, 4.0054e-04, + 3.0518e-04, 1.0925e-02, 1.9775e-02, 1.8921e-02, 1.9165e-02, + 3.7689e-03, 1.3672e-02, 3.6621e-02, 4.9414e-01, -8.8289e-07, + 2.5146e-02, 6.9885e-03, 1.5869e-02, 1.6556e-03, 1.1230e-01, + 2.8312e-06, 2.1338e-05, 3.6478e-05, 1.1719e-02, 1.5076e-02, + 1.5747e-02, 2.3804e-02, 3.1738e-02], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.v_proj.weight': tensor([[-0.0183, 0.0031, 0.0049, ..., 0.0004, -0.0172, -0.0260], + [-0.0047, 0.0093, 0.0052, ..., 0.0046, 0.0053, 0.0051], + [-0.0417, 0.0022, -0.0008, ..., -0.0071, -0.0254, -0.0123], + ..., + [-0.0391, 0.0014, -0.0173, ..., 0.0022, 0.0209, 0.0052], + [ 0.0125, 0.0042, -0.0064, ..., -0.0048, 0.0092, 0.0037], + [ 0.0103, 0.0024, -0.0087, ..., -0.0036, -0.0029, 0.0072]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.output_proj.weight': tensor([[-6.5308e-03, -1.1597e-02, 2.0142e-02, ..., 1.0864e-02, + 1.2268e-02, 5.9509e-03], + [-5.8899e-03, -6.9885e-03, -4.7607e-03, ..., -2.9755e-03, + 5.1270e-03, 1.6724e-02], + [-2.5177e-03, 4.0588e-03, -6.8054e-03, ..., -2.5787e-03, + -2.1210e-03, 3.6926e-03], + ..., + [-3.4790e-03, 1.4954e-03, 6.2561e-03, ..., -6.8054e-03, + 6.3477e-03, 3.5553e-03], + [ 4.3640e-03, 4.1504e-03, 3.9368e-03, ..., -3.3203e-02, + -3.5706e-03, 3.3447e-02], + [ 1.1475e-02, 7.6294e-05, -1.5503e-02, ..., -1.7944e-02, + 3.6011e-03, -4.4556e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_img.weight': tensor([[ 0.0082, -0.0017, 0.0430, ..., -0.0327, -0.0052, 0.0155], + [-0.0278, 0.0280, 0.0249, ..., -0.0072, -0.0378, 0.0271], + [ 0.0134, -0.0121, 0.0312, ..., -0.0077, 0.0253, -0.0113], + ..., + [ 0.0245, -0.0150, 0.0041, ..., -0.0452, 0.0216, -0.0209], + [-0.0081, 0.0069, -0.0471, ..., -0.0106, 0.0157, -0.0082], + [ 0.0062, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0059, -0.0122, -0.0117], + [ 0.0137, 0.0060, -0.0188, ..., 0.0157, 0.0082, -0.0052], + [-0.0306, 0.0012, -0.0272, ..., 0.0100, 0.0337, -0.0050], + ..., + [-0.0040, -0.0046, 0.0044, ..., 0.0209, -0.0107, 0.0071], + [ 0.0193, 0.0210, 0.0123, ..., -0.0146, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0109, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img.weight': tensor([[ 0.0045, -0.0057, 0.0173, ..., 0.0081, 0.0356, 0.0217], + [-0.0498, 0.0062, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0104, 0.0033, -0.0139, ..., -0.0042, 0.0269, -0.0049], + ..., + [ 0.0010, 0.0126, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0073, -0.0162, ..., -0.0057, -0.0073, -0.0120], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0139, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.mlp.layer1.weight': tensor([[-0.0029, 0.0050, -0.0092, ..., 0.0106, 0.0013, -0.0006], + [-0.0026, 0.0053, 0.0047, ..., 0.0059, 0.0116, -0.0123], + [-0.0070, 0.0026, 0.0059, ..., -0.0065, 0.0070, 0.0033], + ..., + [ 0.0194, -0.0036, 0.0075, ..., 0.0078, -0.0006, 0.0038], + [-0.0242, -0.0052, -0.0021, ..., 0.0217, -0.0055, -0.0046], + [ 0.0086, -0.0250, 0.0101, ..., -0.0140, -0.0094, -0.0160]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.mlp.layer2.weight': tensor([[-0.0098, 0.0081, 0.0093, ..., 0.0049, -0.0040, 0.0064], + [-0.0070, 0.0113, 0.0073, ..., -0.0020, 0.0010, -0.0175], + [ 0.0007, 0.0038, -0.0064, ..., 0.0023, 0.0041, -0.0114], + ..., + [-0.0067, -0.0072, 0.0131, ..., -0.0116, -0.0244, 0.0087], + [ 0.0146, 0.0072, -0.0046, ..., -0.0078, 0.0026, -0.0067], + [-0.0046, 0.0062, 0.0068, ..., 0.0039, -0.0120, -0.0143]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0405, -0.0396, -0.0391, ..., 0.0050, 0.0053, 0.0052], + [ 0.0327, 0.0315, 0.0306, ..., -0.0037, -0.0036, -0.0037], + [-0.0417, -0.0398, -0.0396, ..., 0.0050, 0.0051, 0.0051], + ..., + [-0.0447, -0.0435, -0.0422, ..., 0.0034, 0.0036, 0.0037], + [-0.0386, -0.0381, -0.0371, ..., 0.0037, 0.0039, 0.0039], + [-0.0228, -0.0220, -0.0210, ..., 0.0012, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0043, -0.0006, 0.0045, ..., 0.0017, 0.0025, 0.0002], + [-0.0104, 0.0042, -0.0062, ..., -0.0077, -0.0066, -0.0042], + [-0.0049, 0.0018, -0.0061, ..., -0.0084, -0.0084, -0.0020], + ..., + [-0.0220, 0.0037, -0.0267, ..., -0.0286, -0.0270, -0.0038], + [-0.0124, -0.0055, -0.0131, ..., -0.0153, -0.0137, 0.0064], + [ 0.0064, 0.0004, 0.0054, ..., 0.0051, 0.0044, -0.0004]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9570e-02, 5.7861e-02, 5.7129e-02, ..., -1.2159e-04, + -1.6308e-04, -1.1539e-04], + [-4.0283e-03, -3.9978e-03, -3.9978e-03, ..., -4.1008e-05, + -1.1444e-04, -2.2888e-04], + [-1.8555e-02, -1.7944e-02, -1.7090e-02, ..., -6.9809e-04, + -7.0572e-04, -7.1716e-04], + ..., + [-2.9053e-02, -2.8931e-02, -2.8076e-02, ..., -4.1771e-04, + -3.7384e-04, -3.3569e-04], + [-3.3203e-02, -3.2227e-02, -3.0884e-02, ..., -1.8311e-04, + -1.3638e-04, -1.9073e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., -2.0504e-04, + -8.8692e-05, -2.3460e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0089, 0.0040, -0.0204, ..., 0.0052, 0.0135, -0.0127], + [ 0.0073, 0.0063, -0.0064, ..., 0.0061, -0.0039, 0.0053], + [ 0.0121, 0.0049, -0.0063, ..., 0.0080, -0.0047, 0.0102], + ..., + [ 0.0491, 0.0076, -0.0053, ..., 0.0062, -0.0149, 0.0476], + [ 0.0205, -0.0086, 0.0053, ..., -0.0145, -0.0229, 0.0243], + [-0.0029, 0.0096, -0.0164, ..., 0.0114, 0.0080, -0.0063]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_mlp.1.weight': tensor([[-0.0078, -0.0078, -0.0078, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0369, 0.0366, ..., -0.0038, -0.0038, -0.0037], + [ 0.0408, 0.0400, 0.0396, ..., -0.0038, -0.0038, -0.0039], + ..., + [ 0.0262, 0.0253, 0.0254, ..., -0.0021, -0.0021, -0.0020], + [ 0.0057, 0.0063, 0.0064, ..., 0.0014, 0.0014, 0.0013], + [ 0.0166, 0.0170, 0.0167, ..., -0.0025, -0.0025, -0.0026]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_mlp.2.weight': tensor([[-3.2654e-03, 6.0120e-03, 4.9133e-03, ..., 4.3640e-03, + 3.2959e-03, 3.2959e-03], + [ 7.0190e-03, 4.7607e-03, 2.6093e-03, ..., -1.2302e-04, + -7.4158e-03, 3.6240e-05], + [ 3.5553e-03, -4.0894e-03, -5.2490e-03, ..., -3.2806e-03, + -3.6621e-03, -3.2501e-03], + ..., + [ 4.0245e-04, 4.3335e-03, -4.7607e-03, ..., -5.8289e-03, + 3.7994e-03, 3.7537e-03], + [ 7.9956e-03, 1.6968e-02, 1.3916e-02, ..., -4.1504e-03, + -7.9956e-03, -3.8147e-03], + [-4.6387e-03, -4.7913e-03, 2.6855e-03, ..., 4.2114e-03, + 4.6387e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.after_proj.weight': tensor([[ 8.1635e-04, 1.4099e-02, 9.2163e-03, ..., -7.5073e-03, + 9.3994e-03, -5.6458e-03], + [-5.2490e-03, -1.0864e-02, -5.1498e-05, ..., 1.3306e-02, + -3.2959e-03, -6.2866e-03], + [ 3.9673e-03, -7.7820e-03, -8.8120e-04, ..., -5.7068e-03, + -1.7944e-02, 6.0120e-03], + ..., + [-5.3101e-03, 6.7749e-03, 2.5757e-02, ..., -2.0020e-02, + -7.4768e-03, -7.4768e-03], + [-1.1902e-02, -5.5237e-03, 1.6632e-03, ..., 1.1230e-02, + 1.0254e-02, -6.3782e-03], + [-4.4556e-03, 7.5378e-03, -4.5776e-03, ..., -6.2561e-03, + -5.9814e-03, 3.9673e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.after_proj.bias': tensor([ 0.0052, 0.0009, 0.0016, ..., 0.0064, 0.0066, -0.0004], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_proj.weight': tensor([[ 0.0437, -0.0240, -0.0189, ..., 0.0144, -0.0232, 0.0200], + [-0.0255, -0.0161, -0.0065, ..., -0.0056, 0.0101, 0.0036], + [ 0.0311, 0.0232, 0.0126, ..., -0.0045, -0.0199, -0.0047], + ..., + [-0.0085, -0.0048, 0.0018, ..., 0.0041, 0.0041, -0.0118], + [ 0.0167, 0.0259, 0.0025, ..., 0.0038, -0.0092, 0.0112], + [-0.0107, -0.0045, 0.0186, ..., -0.0060, 0.0145, -0.0156]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_norm.weight': tensor([ 1.2344e+00, 3.1836e-01, 4.5117e-01, 2.8711e-01, 1.1172e+00, + 7.7515e-03, 1.1841e-02, -4.2455e-13, 2.0874e-02, 4.4861e-03, + -5.4250e-08, 1.9096e-13, 1.0000e+00, -7.9632e-05, -1.8860e-22, + -2.9419e-02, -6.1035e-03, 8.8501e-03, 1.6235e-02, 1.3916e-02, + 1.5820e-01, -1.4782e-04, 5.0000e-01, 1.3359e+00, 6.4062e-01, + 7.3828e-01, 3.2812e-01, 1.6602e-01, 3.5156e-02, -1.2146e-02, + -3.3875e-03, 5.7459e-05, 5.9509e-03, 1.6785e-04, -6.7749e-03, + 3.3936e-02, 1.1816e-01, -9.1553e-05, 2.4780e-02, 1.1504e-05, + 1.4210e-04, -1.5974e-05, 4.5076e-07, 1.0469e+00, 1.2793e-01, + 1.1875e+00, 6.4062e-01, 1.4355e-01, -4.2969e-02, 1.1108e-02, + 1.0620e-02, 1.6846e-02, 1.3867e-01, -8.9645e-05, 1.6499e-04, + 1.0048e-14, 2.2602e-04, -9.8944e-06, -5.5176e-02, -3.8025e-15, + 3.0640e-02, 7.2718e-06, 5.7983e-03, -3.7231e-03, 5.8594e-01, + 1.1562e+00, 1.1328e+00, 1.1406e+00, 3.1250e-01, -1.2634e-02, + 8.2397e-03, 5.3957e-14, 2.1118e-02, -1.2085e-02, -7.3924e-09, + -5.5067e-14, 4.1504e-02, 2.7008e-03, -3.1764e-20, 6.7188e-01, + 9.1406e-01, -9.2773e-03, -5.6744e-05, 2.6611e-02, 1.1621e-01, + 7.7515e-03, 1.2812e+00, 5.3125e-01, 1.1875e+00, 1.2500e+00, + 2.4121e-01, 4.1992e-01, 4.4922e-02, -1.5869e-02, 4.3335e-03, + 5.0068e-05, -2.5269e-02, -3.8147e-03, 1.7578e-02, 3.5645e-02, + 1.2598e-01, -9.8705e-05, 9.7266e-01, -6.0536e-09, 1.2970e-04, + -1.3471e-05, -1.3279e-10, 6.9531e-01, 1.3750e+00, 1.1953e+00, + 1.2969e+00, 3.3789e-01, 2.5879e-02, 2.0142e-02, 2.1973e-02, + 2.7344e-02, 1.3574e-01, -1.0347e-04, 1.6594e-04, 1.6542e-14, + 1.5068e-04, -1.1206e-05, 1.0469e+00, -4.8572e-15, 3.5889e-02, + 1.0058e-06, 5.9375e-01, 1.0547e+00], dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.self_attn.k_proj.weight': tensor([[-0.0256, 0.0096, 0.0081, ..., 0.0226, -0.0024, 0.0109], + [ 0.0042, 0.0034, 0.0022, ..., -0.0066, 0.0105, -0.0008], + [-0.0291, 0.0134, -0.0076, ..., -0.0049, 0.0238, 0.0005], + ..., + [-0.0023, -0.0036, 0.0055, ..., -0.0068, -0.0052, 0.0079], + [ 0.0024, 0.0055, 0.0126, ..., 0.0192, -0.0034, -0.0075], + [ 0.0066, 0.0018, 0.0126, ..., -0.0008, 0.0061, 0.0229]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.k_norm.weight': tensor([ 1.2812e+00, 7.1875e-01, 4.3555e-01, 2.9297e-01, 1.1172e+00, + 1.2939e-02, 1.9653e-02, 2.4158e-13, 2.1484e-02, 7.6904e-03, + -6.1002e-08, 1.9895e-13, 1.0000e+00, -7.9632e-05, -7.0310e-23, + -7.6904e-03, 2.6855e-03, 8.0078e-02, 1.8066e-02, 3.1586e-03, + 5.7678e-03, -1.0490e-04, 5.3125e-01, 1.2188e+00, 2.1191e-01, + 5.8594e-01, 4.0430e-01, 1.5625e-01, 1.0193e-02, -1.8692e-03, + 4.6997e-03, 4.8161e-05, 4.1199e-03, 4.0283e-02, 4.8218e-03, + -4.3640e-03, 4.3335e-03, -3.9816e-05, 2.9907e-02, 1.1981e-05, + 1.6880e-04, -1.2636e-05, -4.5449e-07, 1.1406e+00, 7.5781e-01, + 1.1406e+00, 4.4141e-01, 2.2070e-01, 1.6846e-02, 5.6641e-02, + -9.8267e-03, 2.8992e-03, 1.1414e-02, -2.9683e-05, 6.2943e-05, + -1.3600e-14, 2.4605e-04, -3.6210e-06, -2.2217e-02, -3.5388e-15, + -3.8452e-03, -4.6492e-06, 3.6621e-03, 1.3733e-02, 6.2891e-01, + 1.0859e+00, 1.1484e+00, 1.1172e+00, 2.8516e-01, 1.8311e-03, + -2.2125e-03, -2.4691e-13, 1.6357e-02, 1.8539e-03, -7.6252e-09, + -4.5075e-14, -1.3184e-02, 2.7008e-03, 3.1764e-20, 6.7188e-01, + 9.1406e-01, 6.5430e-02, 1.5926e-04, 3.8147e-03, 8.8501e-03, + 7.9956e-03, 1.2109e+00, 7.9688e-01, 1.3047e+00, 1.2969e+00, + 2.5146e-02, 4.2188e-01, 1.7578e-02, 2.1362e-02, -3.4027e-03, + 3.6001e-05, 2.7954e-02, 4.5410e-02, 2.3193e-02, -8.1177e-03, + -1.8311e-02, -6.0320e-05, 9.7266e-01, 2.7707e-08, 1.6117e-04, + -9.1195e-06, 7.1907e-12, 4.7070e-01, 1.2188e+00, 1.1484e+00, + 1.4141e+00, 3.3789e-01, -3.8330e-02, 1.0205e-01, 2.1729e-02, + 1.2024e-02, 1.1475e-02, -5.6267e-05, 1.5354e-04, -2.2871e-14, + 1.7929e-04, -6.0499e-06, 1.0469e+00, -5.3291e-15, -1.5625e-02, + 1.2964e-06, 5.9375e-01, 1.0547e+00], dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0095, -0.0096, ..., -0.0177, 0.0064, 0.0099], + [-0.0042, -0.0049, -0.0126, ..., -0.0092, 0.0050, 0.0063], + [-0.0137, -0.0091, 0.0016, ..., 0.0320, -0.0079, -0.0100], + ..., + [ 0.0110, 0.0151, -0.0223, ..., -0.0342, -0.0092, -0.0047], + [ 0.0007, 0.0009, 0.0025, ..., 0.0045, 0.0035, -0.0098], + [ 0.0040, -0.0181, -0.0162, ..., 0.0044, -0.0074, -0.0038]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.output_proj.weight': tensor([[ 0.0148, 0.0173, -0.0488, ..., 0.0023, -0.0142, 0.0139], + [-0.0038, -0.0066, -0.0146, ..., 0.0210, 0.0125, -0.0125], + [ 0.0003, 0.0044, 0.0043, ..., 0.0037, 0.0005, 0.0061], + ..., + [ 0.0070, 0.0051, -0.0271, ..., 0.0037, 0.0056, 0.0059], + [ 0.0009, 0.0048, 0.0065, ..., 0.0038, 0.0160, -0.0135], + [ 0.0040, 0.0108, 0.0039, ..., -0.0024, 0.0085, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_proj.weight': tensor([[-0.0053, 0.0062, -0.0029, ..., 0.0040, 0.0022, 0.0024], + [ 0.0075, -0.0033, 0.0025, ..., 0.0078, 0.0012, 0.0060], + [ 0.0046, 0.0026, -0.0014, ..., -0.0047, 0.0080, -0.0036], + ..., + [-0.0021, 0.0029, -0.0014, ..., -0.0053, 0.0069, -0.0017], + [-0.0088, -0.0069, 0.0020, ..., 0.0078, -0.0011, 0.0081], + [-0.0026, -0.0013, -0.0044, ..., -0.0045, 0.0035, -0.0061]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_norm.weight': tensor([ 2.8076e-03, 1.4062e-01, 2.1118e-02, 6.1719e-01, 2.1240e-02, + 1.4526e-02, 1.2695e-02, 5.3406e-03, 2.0386e-02, 1.2305e-01, + 1.7212e-02, 5.6641e-02, 2.6758e-01, 2.2217e-02, 1.7969e-01, + 2.2461e-02, 6.3965e-02, 2.4292e-02, 5.4016e-03, 2.3438e-02, + 1.0681e-02, 1.6895e-01, 2.6562e-01, 2.0410e-01, 2.4902e-02, + 1.5820e-01, 5.5420e-02, 1.1475e-02, 3.8867e-01, 1.2695e-02, + 1.5015e-02, 5.0781e-01, 4.6631e-02, 4.3701e-02, 9.1553e-03, + 2.0020e-02, 8.9355e-02, 5.0391e-01, 5.7422e-01, 1.1841e-02, + 1.1414e-02, 1.0059e-01, 5.2795e-03, 5.5469e-01, 2.9297e-02, + 4.3213e-02, 1.1169e-02, 3.9648e-01, 8.3618e-03, 2.6978e-02, + 6.7871e-02, 1.5747e-02, 2.0264e-02, 5.1953e-01, 2.7313e-03, + 2.8076e-02, 4.7119e-02, 6.7871e-02, 1.1719e-02, 2.0752e-02, + 3.1281e-03, 1.4587e-02, 5.2490e-03, 3.6328e-01, 2.3193e-02, + 1.7700e-03, 2.4902e-02, 6.4453e-02, 5.5664e-02, 9.1553e-03, + 9.4604e-03, 6.2109e-01, -3.6716e-05, 6.0547e-01, 2.0386e-02, + 4.1962e-04, 1.4587e-02, 3.1055e-01, 2.9297e-02, 4.5586e-04, + 8.1055e-02, 8.2397e-03, 5.2246e-02, 9.8877e-03, 1.6235e-02, + 8.0566e-03, 1.1719e-02, 5.1025e-02, 2.0996e-01, 1.1536e-02, + 1.0437e-02, 2.3560e-02, 4.2480e-02, 3.3447e-02, 3.0640e-02, + 1.7456e-02, 1.3550e-02, 2.3560e-02, 5.0537e-02, 2.7466e-02, + 5.0049e-02, 1.9897e-02, 6.2561e-03, 5.2246e-02, 5.9766e-01, + 6.0547e-02, 7.2021e-03, 4.6875e-01, 1.5015e-02, 2.8931e-02, + 1.2085e-02, 1.4587e-02, 7.8201e-05, 7.9102e-02, 4.5117e-01, + 2.0264e-02, 2.4048e-02, 5.6641e-01, 1.0254e-01, 3.1128e-02, + 1.5015e-02, 6.9531e-01, 1.4062e-01, 1.6235e-02, 1.1169e-02, + 7.8125e-03, 2.2461e-02, 1.2634e-02], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_proj.weight': tensor([[ 0.0013, -0.0034, 0.0126, ..., -0.0002, -0.0052, 0.0074], + [-0.0046, 0.0003, -0.0056, ..., 0.0067, -0.0096, -0.0072], + [-0.0070, -0.0016, -0.0056, ..., 0.0168, -0.0151, -0.0118], + ..., + [-0.0018, 0.0046, -0.0001, ..., -0.0026, -0.0044, -0.0068], + [ 0.0043, 0.0029, -0.0074, ..., -0.0056, 0.0054, 0.0016], + [-0.0072, 0.0007, 0.0049, ..., 0.0015, 0.0039, 0.0058]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_norm.weight': tensor([ 2.8381e-03, 1.4062e-01, 2.0996e-02, 6.1719e-01, 2.0996e-02, + 1.4526e-02, 1.2817e-02, 5.2185e-03, 2.0630e-02, 1.2305e-01, + 1.7212e-02, 5.5176e-02, 2.6758e-01, 2.2095e-02, 1.7969e-01, + 2.3193e-02, 6.3477e-02, 2.4414e-02, 5.3406e-03, 2.3315e-02, + 1.0986e-02, 1.6992e-01, 2.6758e-01, 2.0410e-01, 2.4780e-02, + 1.5820e-01, 5.4443e-02, 9.3994e-03, 3.9062e-01, 1.2817e-02, + 1.5198e-02, 5.0781e-01, 4.6631e-02, 4.3701e-02, 9.5215e-03, + 1.9653e-02, 8.8867e-02, 5.0391e-01, 5.7422e-01, 1.1780e-02, + 1.1414e-02, 1.0010e-01, 5.2490e-03, 5.5469e-01, 2.9541e-02, + 4.3213e-02, 1.1475e-02, 3.9648e-01, 8.4839e-03, 2.6855e-02, + 6.7871e-02, 1.6357e-02, 2.0386e-02, 5.1953e-01, 3.1128e-03, + 2.7954e-02, 4.6387e-02, 6.7871e-02, 1.1780e-02, 1.9287e-02, + 3.0060e-03, 1.4526e-02, 5.6152e-03, 3.6328e-01, 2.2949e-02, + 1.9302e-03, 2.4414e-02, 6.4453e-02, 5.5664e-02, 9.0942e-03, + 9.8877e-03, 6.2109e-01, -5.1498e-05, 6.0547e-01, 2.0508e-02, + -1.7548e-04, 1.4648e-02, 3.0859e-01, 3.0029e-02, 6.9046e-04, + 8.1543e-02, 8.1787e-03, 5.1514e-02, 9.7656e-03, 1.5747e-02, + 8.1787e-03, 1.2329e-02, 5.1758e-02, 2.0898e-01, 1.1414e-02, + 1.0498e-02, 2.2583e-02, 4.2725e-02, 3.3203e-02, 3.1494e-02, + 1.7456e-02, 1.3611e-02, 2.3560e-02, 5.0537e-02, 2.7344e-02, + 4.9561e-02, 2.0020e-02, 6.8054e-03, 5.2734e-02, 5.9766e-01, + 6.0547e-02, 7.1411e-03, 4.6875e-01, 1.4587e-02, 2.8809e-02, + 1.2756e-02, 1.4587e-02, 8.0109e-05, 7.9102e-02, 4.5117e-01, + 1.9897e-02, 2.4048e-02, 5.6641e-01, 1.0352e-01, 3.1250e-02, + 1.4771e-02, 6.9922e-01, 1.3965e-01, 1.5259e-02, 1.0803e-02, + 7.6599e-03, 2.2339e-02, 1.2390e-02], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.v_proj.weight': tensor([[ 0.0130, -0.0059, -0.0171, ..., 0.0042, -0.0138, -0.0076], + [-0.0179, -0.0020, -0.0033, ..., -0.0018, -0.0227, 0.0190], + [-0.0127, -0.0049, 0.0033, ..., -0.0045, 0.0310, 0.0281], + ..., + [-0.0030, -0.0030, -0.0121, ..., 0.0048, 0.0171, 0.0043], + [ 0.0278, 0.0032, 0.0175, ..., 0.0035, 0.0116, 0.0164], + [ 0.0128, -0.0001, 0.0049, ..., 0.0053, -0.0192, -0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.output_proj.weight': tensor([[ 0.0125, -0.0001, -0.0154, ..., -0.0190, 0.0078, -0.0012], + [ 0.0089, 0.0038, 0.0125, ..., 0.0066, -0.0090, 0.0024], + [ 0.0059, -0.0033, -0.0007, ..., 0.0007, -0.0078, -0.0057], + ..., + [-0.0114, 0.0038, 0.0079, ..., 0.0052, -0.0232, 0.0073], + [ 0.0097, -0.0165, 0.0049, ..., -0.0265, -0.0133, -0.0056], + [ 0.0042, -0.0167, 0.0177, ..., 0.0317, -0.0025, 0.0086]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0065, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [-0.0032, 0.0150, -0.0159, ..., -0.0039, -0.0054, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0071, -0.0067], + [-0.0337, -0.0040, 0.0054, ..., -0.0041, 0.0172, 0.0231], + [-0.0255, -0.0040, -0.0039, ..., 0.0408, 0.0430, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0017], + [ 0.0198, -0.0208, 0.0128, ..., 0.0090, -0.0337, -0.0050], + [ 0.0302, 0.0101, -0.0029, ..., -0.0304, 0.0199, 0.0045], + ..., + [ 0.0171, -0.0276, -0.0073, ..., 0.0231, 0.0182, -0.0040], + [ 0.0221, 0.0388, 0.0056, ..., 0.0023, -0.0488, -0.0361], + [ 0.0058, 0.0118, -0.0055, ..., 0.0042, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img.weight': tensor([[ 2.7344e-02, -1.9908e-05, 1.1841e-02, ..., 1.4725e-03, + 3.4668e-02, 6.1523e-02], + [-2.5757e-02, 2.6703e-03, -1.9653e-02, ..., 1.3916e-02, + -4.6082e-03, 1.0986e-02], + [-8.2397e-03, -1.6846e-02, -4.2725e-02, ..., -5.8594e-02, + 1.5320e-02, 8.5449e-03], + ..., + [ 9.5825e-03, 1.3184e-02, 1.8433e-02, ..., 1.1047e-02, + 8.9111e-03, 5.1514e-02], + [-3.2806e-03, -2.4658e-02, -2.7832e-02, ..., -8.9722e-03, + -3.8818e-02, 8.3618e-03], + [ 1.9043e-02, -2.6733e-02, -2.1973e-02, ..., -4.5654e-02, + -1.5991e-02, 1.6113e-02]], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.mlp.layer1.weight': tensor([[-0.0068, 0.0051, 0.0058, ..., 0.0002, -0.0018, -0.0094], + [-0.0115, 0.0060, 0.0082, ..., 0.0094, -0.0030, 0.0022], + [ 0.0054, -0.0074, 0.0063, ..., 0.0050, -0.0054, -0.0349], + ..., + [-0.0066, -0.0027, 0.0054, ..., 0.0137, 0.0046, -0.0133], + [ 0.0049, 0.0027, -0.0034, ..., -0.0194, -0.0060, -0.0376], + [ 0.0045, 0.0277, 0.0002, ..., -0.0041, 0.0078, -0.0072]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.mlp.layer2.weight': tensor([[ 0.0056, 0.0057, -0.0063, ..., -0.0060, 0.0172, -0.0059], + [-0.0040, -0.0027, -0.0060, ..., 0.0073, -0.0140, -0.0192], + [ 0.0028, 0.0055, 0.0098, ..., 0.0089, 0.0079, -0.0041], + ..., + [-0.0043, 0.0032, -0.0209, ..., -0.0107, 0.0076, -0.0089], + [-0.0117, 0.0106, 0.0123, ..., -0.0101, 0.0023, -0.0024], + [-0.0011, 0.0063, -0.0120, ..., 0.0049, -0.0094, -0.0374]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0347, 0.0344, 0.0330, ..., -0.0041, -0.0042, -0.0043], + [-0.0334, -0.0322, -0.0317, ..., 0.0043, 0.0047, 0.0046], + [-0.0349, -0.0344, -0.0332, ..., 0.0045, 0.0048, 0.0048], + ..., + [-0.0393, -0.0381, -0.0374, ..., 0.0022, 0.0023, 0.0024], + [-0.0400, -0.0393, -0.0383, ..., 0.0021, 0.0022, 0.0022], + [ 0.0078, 0.0078, 0.0073, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0079, -0.0075, -0.0080, ..., -0.0095, -0.0101, 0.0018], + [ 0.0047, -0.0041, -0.0047, ..., -0.0023, -0.0013, 0.0017], + [ 0.0089, -0.0084, -0.0087, ..., -0.0045, -0.0043, -0.0007], + ..., + [ 0.0383, -0.0422, -0.0413, ..., -0.0222, -0.0182, 0.0082], + [ 0.0093, -0.0092, -0.0087, ..., -0.0067, -0.0067, -0.0113], + [-0.0181, 0.0188, 0.0179, ..., 0.0146, 0.0138, 0.0118]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9326e-02, 5.8350e-02, 5.7373e-02, ..., -1.2589e-04, + -5.5313e-05, -2.0790e-04], + [ 6.0547e-02, 5.8838e-02, 5.7129e-02, ..., 3.7575e-04, + 4.8828e-04, 5.1880e-04], + [-4.9561e-02, -4.7852e-02, -4.6631e-02, ..., -1.2934e-05, + 4.2439e-05, -6.6280e-05], + ..., + [ 3.5156e-02, 3.4180e-02, 3.3447e-02, ..., 1.0538e-04, + 3.7432e-05, 6.6757e-05], + [ 6.2988e-02, 6.2012e-02, 6.0791e-02, ..., 5.6076e-04, + 5.9891e-04, 5.9128e-04], + [-5.1758e-02, -5.1270e-02, -5.0049e-02, ..., -3.5286e-04, + -3.9864e-04, -3.5858e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0064, 0.0097, -0.0022, ..., -0.0041, -0.0103, 0.0048], + [-0.0140, 0.0099, 0.0109, ..., -0.0084, -0.0104, 0.0062], + [ 0.0036, 0.0033, -0.0029, ..., 0.0036, -0.0033, 0.0033], + ..., + [ 0.0182, 0.0115, -0.0118, ..., 0.0065, 0.0120, 0.0056], + [ 0.0114, 0.0291, -0.0128, ..., 0.0134, 0.0383, -0.0232], + [-0.0200, 0.0195, 0.0177, ..., -0.0133, -0.0099, 0.0231]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_mlp.1.weight': tensor([[-0.0757, -0.0742, -0.0718, ..., 0.0029, 0.0032, 0.0031], + [-0.0258, -0.0243, -0.0233, ..., 0.0006, 0.0007, 0.0006], + [ 0.0559, 0.0542, 0.0530, ..., -0.0017, -0.0017, -0.0017], + ..., + [ 0.0430, 0.0420, 0.0415, ..., -0.0019, -0.0018, -0.0016], + [ 0.0347, 0.0330, 0.0317, ..., -0.0009, -0.0008, -0.0009], + [-0.0520, -0.0505, -0.0496, ..., 0.0022, 0.0022, 0.0021]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_mlp.2.weight': tensor([[-0.0266, -0.0076, 0.0162, ..., 0.0093, 0.0091, -0.0166], + [ 0.0042, -0.0098, -0.0094, ..., -0.0092, 0.0098, 0.0092], + [-0.0038, -0.0087, -0.0049, ..., -0.0069, 0.0090, 0.0049], + ..., + [ 0.0608, -0.0090, -0.0262, ..., -0.0195, 0.0083, 0.0270], + [ 0.0187, -0.0172, -0.0186, ..., -0.0171, 0.0171, 0.0179], + [ 0.0054, -0.0179, -0.0103, ..., -0.0129, 0.0184, 0.0098]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.after_proj.weight': tensor([[-0.0054, 0.0061, -0.0078, ..., -0.0073, -0.0055, -0.0062], + [-0.0144, -0.0081, 0.0093, ..., 0.0078, -0.0068, 0.0075], + [ 0.0061, 0.0078, 0.0070, ..., -0.0076, 0.0082, -0.0051], + ..., + [ 0.0074, -0.0090, 0.0143, ..., 0.0154, 0.0192, 0.0157], + [ 0.0014, 0.0089, 0.0025, ..., -0.0200, -0.0118, 0.0017], + [-0.0080, -0.0099, -0.0138, ..., -0.0076, -0.0044, -0.0057]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.after_proj.bias': tensor([ 0.0055, -0.0029, 0.0007, ..., 0.0034, 0.0079, 0.0055], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_proj.weight': tensor([[-0.0192, 0.0198, 0.0232, ..., -0.0031, -0.0097, -0.0026], + [-0.0106, -0.0056, -0.0008, ..., 0.0075, -0.0066, -0.0070], + [-0.0182, -0.0057, 0.0047, ..., -0.0069, -0.0079, 0.0070], + ..., + [-0.0043, 0.0033, -0.0081, ..., -0.0080, 0.0104, 0.0029], + [ 0.0066, -0.0008, -0.0037, ..., -0.0289, -0.0199, -0.0289], + [ 0.0036, -0.0161, 0.0070, ..., 0.0107, -0.0027, -0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_norm.weight': tensor([ 9.3750e-01, 1.4609e+00, 1.3750e+00, 1.3594e+00, 1.3574e-01, + 5.8350e-02, 1.6479e-02, -2.1770e-08, -2.4435e-33, 3.0823e-03, + 5.4016e-03, 1.0840e-01, -6.6223e-03, -1.0315e-02, 1.9989e-03, + 1.4687e-04, 5.9326e-02, 9.1406e-01, 1.0156e+00, 3.9291e-04, + 1.4400e-04, 1.1406e+00, 1.6562e+00, 1.6406e+00, 1.0000e+00, + 8.8672e-01, 1.2969e+00, -1.0376e-02, 4.2439e-05, 6.1989e-05, + -9.3994e-03, 2.6131e-04, 2.0874e-02, 1.0078e+00, 9.3384e-03, + 1.0312e+00, 2.8076e-03, 8.0109e-05, 2.3438e-02, -8.1177e-03, + 5.2527e-07, 1.2207e-04, 1.9908e-05, 1.6797e+00, 1.2266e+00, + 1.2734e+00, 1.3438e+00, 1.4531e+00, 6.7871e-02, 1.7452e-04, + 2.9053e-02, 2.6172e-01, 1.0469e+00, 9.0820e-02, 5.1737e-05, + 1.0596e-01, 2.3804e-03, 1.1621e-01, 2.5630e-05, 6.9580e-03, + 1.9372e-06, -1.8845e-03, 3.7231e-03, 1.0938e+00, 1.5234e+00, + 6.7969e-01, 8.3594e-01, 1.1406e+00, 6.0303e-02, -3.1738e-02, + 3.4637e-03, -4.9127e-08, -3.9482e-33, 2.0752e-02, 1.4709e-02, + 2.5482e-03, 3.1281e-04, -2.6093e-03, -1.6724e-02, 1.3062e-02, + 5.8594e-02, -2.8076e-03, 1.1353e-02, 2.4986e-04, 1.5640e-04, + -6.5002e-03, 6.8359e-01, 6.7578e-01, 1.3281e+00, 1.3438e+00, + 7.0312e-01, 1.0681e-03, -3.3140e-05, 3.1471e-05, 4.6387e-03, + -4.5967e-04, 2.9297e-02, 1.6113e-01, 9.0332e-03, 6.4453e-02, + 1.9932e-04, 3.4809e-05, 1.2207e-01, -7.6771e-05, 5.1498e-05, + -2.6226e-05, 5.1260e-05, 4.9805e-01, 1.4453e+00, 1.3047e+00, + 8.4375e-01, 8.4375e-01, 7.3730e-02, -1.7166e-03, 2.7832e-02, + 1.0547e+00, 5.9570e-02, 9.8047e-01, 4.8161e-05, 2.8839e-03, + 2.6245e-02, 5.2490e-02, 1.2815e-05, 5.5695e-04, 9.3937e-05, + 2.3804e-03, 1.0156e+00, -1.0071e-02], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.self_attn.k_proj.weight': tensor([[-0.0176, -0.0067, -0.0075, ..., -0.0074, 0.0067, -0.0098], + [-0.0050, -0.0165, 0.0178, ..., -0.0085, -0.0135, 0.0146], + [-0.0064, 0.0046, 0.0004, ..., -0.0054, -0.0054, 0.0057], + ..., + [-0.0082, 0.0190, -0.0006, ..., 0.0033, 0.0077, 0.0007], + [ 0.0146, -0.0021, 0.0188, ..., -0.0072, 0.0193, -0.0062], + [-0.0028, -0.0042, 0.0037, ..., -0.0037, 0.0043, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.k_norm.weight': tensor([ 7.0312e-01, 1.4141e+00, 1.3672e+00, 1.3047e+00, 1.3672e-01, + 6.5918e-02, 1.1108e-02, -1.8976e-08, -1.8296e-33, -6.6376e-04, + 5.2185e-03, 1.0938e-01, -1.0681e-02, 3.4668e-02, -1.7929e-03, + 5.9843e-05, 3.9673e-03, 9.1406e-01, 1.0156e+00, 4.6921e-04, + 8.9645e-05, 1.1406e+00, 1.4844e+00, 1.4844e+00, 1.6235e-02, + 2.8516e-01, 1.3281e+00, 1.9302e-03, 4.5538e-05, -5.1498e-05, + -8.7280e-03, 5.8746e-04, -1.9287e-02, 1.0000e+00, 6.2561e-03, + 1.0234e+00, 2.6855e-03, 8.8692e-05, -2.5146e-02, -8.1787e-03, + -1.0505e-06, -1.2589e-04, 1.0550e-05, 1.3203e+00, 5.5859e-01, + 1.2031e+00, 1.3984e+00, 1.4297e+00, -4.9438e-03, -7.7820e-04, + 2.4719e-03, -2.2095e-02, 1.0469e+00, 6.2256e-02, -3.4809e-05, + 1.0596e-01, -1.2970e-03, 1.1621e-01, -7.5400e-06, 7.0801e-02, + -1.7583e-06, 2.1606e-02, -2.1484e-02, 1.0938e+00, 1.5625e+00, + 1.0312e+00, 7.8125e-01, 1.1484e+00, 4.9072e-02, -1.9775e-02, + -1.8799e-02, 5.2620e-08, 4.7426e-33, 2.1240e-02, 2.5391e-02, + 3.3936e-02, 7.2098e-04, 2.7344e-02, -3.0151e-02, -1.3123e-02, + -2.6855e-02, 2.8381e-03, 2.7618e-03, 3.1853e-04, 9.2506e-05, + 5.7678e-03, 1.1875e+00, 1.1484e+00, 1.4297e+00, 1.4141e+00, + 3.0859e-01, 1.1536e-02, -6.0320e-05, -4.3631e-05, 6.0425e-03, + -2.2507e-04, 1.2695e-02, 3.2806e-03, -8.5449e-03, -2.2095e-02, + 9.9182e-05, 5.6267e-05, 1.2158e-01, 3.9339e-05, 6.2943e-05, + 2.7657e-05, 4.7445e-05, 1.2344e+00, 1.6484e+00, 1.2734e+00, + 2.3730e-01, 7.7734e-01, -6.0120e-03, -1.5182e-03, -1.5320e-02, + 1.0547e+00, -2.9053e-02, 9.8438e-01, 4.4823e-05, 7.5989e-03, + 2.6489e-02, 5.2002e-02, 8.0466e-06, 4.1748e-02, 9.4414e-05, + 1.5015e-02, 1.0156e+00, -1.9302e-03], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.self_attn.v_proj.weight': tensor([[ 3.8147e-03, -6.7444e-03, 1.0315e-02, ..., 4.6631e-02, + -9.6893e-04, 2.5330e-03], + [ 1.1963e-02, 2.1973e-02, -8.9722e-03, ..., -2.1240e-02, + -8.8501e-03, -5.7678e-03], + [-2.0294e-03, 1.8677e-02, -4.8218e-03, ..., 1.5625e-02, + -4.4556e-03, 9.4604e-03], + ..., + [ 4.2200e-05, -4.1199e-03, -1.9531e-02, ..., 9.4604e-03, + 1.5381e-02, 3.2349e-03], + [ 2.8229e-03, 1.8433e-02, -5.8899e-03, ..., 3.9978e-03, + 4.2114e-03, -3.4943e-03], + [-9.0332e-03, -8.6060e-03, 5.1880e-03, ..., -6.7139e-03, + -8.2397e-03, -7.4463e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.output_proj.weight': tensor([[ 0.0072, 0.0190, 0.0056, ..., 0.0101, 0.0192, -0.0095], + [ 0.0275, 0.0043, -0.0049, ..., -0.0041, -0.0270, -0.0094], + [ 0.0039, 0.0334, -0.0115, ..., -0.0043, -0.0165, 0.0060], + ..., + [-0.0381, -0.0177, 0.0038, ..., 0.0108, -0.0082, -0.0183], + [ 0.0012, 0.0125, -0.0266, ..., -0.0176, 0.0238, -0.0033], + [-0.0432, 0.0046, 0.0233, ..., 0.0222, -0.0132, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_proj.weight': tensor([[-0.0003, -0.0081, -0.0045, ..., -0.0159, -0.0014, 0.0152], + [ 0.0070, -0.0017, -0.0037, ..., 0.0017, 0.0044, 0.0084], + [ 0.0084, 0.0048, -0.0092, ..., 0.0050, 0.0126, -0.0039], + ..., + [ 0.0060, 0.0171, 0.0045, ..., 0.0056, -0.0106, 0.0049], + [-0.0036, 0.0073, -0.0050, ..., -0.0056, -0.0114, 0.0037], + [ 0.0004, -0.0054, 0.0040, ..., -0.0049, 0.0120, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_norm.weight': tensor([0.1113, 0.0237, 0.4238, 0.0752, 0.1670, 0.0310, 0.1553, 0.1206, 0.5195, + 0.0447, 0.0654, 0.0161, 0.0518, 0.0674, 0.0640, 0.2021, 0.4570, 0.0344, + 0.7539, 0.0364, 0.0327, 0.1689, 0.0193, 0.0894, 0.0623, 0.4355, 0.5234, + 0.2754, 0.0767, 0.0349, 0.0139, 0.2305, 0.6719, 0.2539, 0.0466, 0.0271, + 0.0216, 0.0967, 0.0184, 0.0383, 0.0527, 0.1201, 0.0229, 0.0439, 0.2080, + 0.0498, 0.3555, 0.0260, 0.0554, 0.0211, 0.0258, 0.4023, 0.5312, 0.3789, + 0.1025, 0.5898, 0.5938, 0.0366, 0.1367, 0.0144, 0.0913, 0.4805, 0.0153, + 0.5273, 0.1699, 0.4160, 0.1016, 0.0498, 0.0732, 0.4277, 0.0281, 0.5508, + 0.0251, 0.0442, 0.1533, 0.0908, 0.0243, 0.1113, 0.6406, 0.0398, 0.0952, + 0.0879, 0.0292, 0.0461, 0.4219, 0.0674, 0.0708, 0.0854, 0.0172, 0.1533, + 0.0432, 0.1787, 0.0471, 0.0703, 0.0391, 0.1201, 0.0114, 0.3555, 0.5391, + 0.0503, 0.0549, 0.3242, 0.0339, 0.1309, 0.0398, 0.2285, 0.0156, 0.5742, + 0.5273, 0.0359, 0.0171, 0.7344, 0.0273, 0.0225, 0.4746, 0.2178, 0.3008, + 0.7227, 0.0369, 0.1001, 0.0801, 0.5625, 0.6055, 0.0199, 0.0188, 0.1914, + 0.0374, 0.0408], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_proj.weight': tensor([[-1.9531e-03, 1.8158e-03, 6.0730e-03, ..., -1.7090e-03, + 1.0925e-02, -1.0864e-02], + [ 5.8289e-03, -6.7711e-05, -1.7776e-03, ..., 1.7548e-03, + -3.8452e-03, 1.2390e-02], + [-1.0437e-02, 7.6904e-03, -1.3428e-02, ..., 1.0803e-02, + 5.3406e-03, 4.2114e-03], + ..., + [ 7.5378e-03, 3.1128e-03, -7.1411e-03, ..., -4.4250e-04, + -1.6724e-02, 5.8594e-03], + [-8.6060e-03, -1.8539e-03, -6.6833e-03, ..., 2.8381e-03, + -4.6692e-03, 1.7853e-03], + [ 9.9487e-03, -1.4038e-03, 5.2795e-03, ..., -7.4768e-03, + 7.4463e-03, -5.7983e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_norm.weight': tensor([0.1118, 0.0237, 0.4238, 0.0752, 0.1670, 0.0317, 0.1553, 0.1206, 0.5195, + 0.0452, 0.0664, 0.0161, 0.0518, 0.0674, 0.0640, 0.2041, 0.4590, 0.0344, + 0.7500, 0.0361, 0.0325, 0.1699, 0.0194, 0.0889, 0.0618, 0.4316, 0.5234, + 0.2734, 0.0767, 0.0349, 0.0150, 0.2324, 0.6719, 0.2539, 0.0474, 0.0271, + 0.0214, 0.0967, 0.0186, 0.0378, 0.0525, 0.1211, 0.0233, 0.0444, 0.2080, + 0.0498, 0.3555, 0.0262, 0.0557, 0.0208, 0.0262, 0.4023, 0.5312, 0.3789, + 0.1025, 0.5898, 0.5977, 0.0366, 0.1367, 0.0147, 0.0913, 0.4805, 0.0156, + 0.5273, 0.1689, 0.4160, 0.1011, 0.0498, 0.0732, 0.4297, 0.0282, 0.5508, + 0.0249, 0.0439, 0.1533, 0.0903, 0.0243, 0.1118, 0.6406, 0.0398, 0.0942, + 0.0879, 0.0292, 0.0461, 0.4219, 0.0679, 0.0698, 0.0850, 0.0175, 0.1553, + 0.0442, 0.1777, 0.0471, 0.0708, 0.0393, 0.1206, 0.0114, 0.3555, 0.5391, + 0.0498, 0.0552, 0.3242, 0.0339, 0.1309, 0.0398, 0.2285, 0.0156, 0.5742, + 0.5273, 0.0356, 0.0171, 0.7344, 0.0266, 0.0227, 0.4766, 0.2178, 0.3008, + 0.7188, 0.0364, 0.0981, 0.0801, 0.5625, 0.6055, 0.0198, 0.0188, 0.1914, + 0.0386, 0.0410], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.v_proj.weight': tensor([[ 0.0084, -0.0016, -0.0292, ..., 0.0041, 0.0038, -0.0220], + [ 0.0032, 0.0095, -0.0042, ..., 0.0006, -0.0060, -0.0199], + [ 0.0332, -0.0038, -0.0092, ..., -0.0049, -0.0175, -0.0386], + ..., + [-0.0118, 0.0019, 0.0160, ..., -0.0067, -0.0212, 0.0049], + [-0.0311, -0.0029, -0.0271, ..., -0.0032, 0.0128, -0.0242], + [-0.0053, 0.0071, 0.0076, ..., 0.0010, 0.0092, 0.0043]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0059, -0.0087, ..., -0.0060, 0.0027, 0.0042], + [-0.0173, -0.0145, -0.0193, ..., 0.0061, -0.0162, -0.0192], + [ 0.0070, -0.0029, -0.0039, ..., -0.0020, -0.0012, -0.0052], + ..., + [ 0.0082, -0.0060, 0.0193, ..., 0.0190, -0.0111, -0.0039], + [-0.0043, -0.0080, -0.0049, ..., 0.0244, -0.0148, 0.0232], + [-0.0089, -0.0129, 0.0269, ..., -0.0228, 0.0344, 0.0085]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0049, -0.0066, ..., 0.0266, -0.0278, 0.0076], + ..., + [-0.0217, -0.0410, 0.0488, ..., 0.0013, -0.0027, -0.0255], + [ 0.0015, 0.0066, 0.0024, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0057, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, 0.0027, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0033, 0.0293, ..., 0.0066, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0139, 0.0126, -0.0092], + ..., + [-0.0043, 0.0320, -0.0233, ..., -0.0067, 0.0016, 0.0188], + [ 0.0134, -0.0352, 0.0267, ..., -0.0305, -0.0021, 0.0037], + [-0.0205, 0.0140, -0.0019, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0019, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0057, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0016], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, -0.0013, 0.0015], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0022, ..., -0.0103, 0.0073, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.mlp.layer1.weight': tensor([[ 0.0108, 0.0060, -0.0140, ..., 0.0150, 0.0309, -0.0071], + [-0.0115, 0.0194, 0.0139, ..., -0.0099, -0.0183, 0.0038], + [-0.0079, 0.0284, -0.0168, ..., 0.0104, -0.0039, 0.0002], + ..., + [ 0.0049, 0.0046, 0.0630, ..., 0.0175, 0.0330, 0.0019], + [ 0.0021, -0.0042, -0.0120, ..., -0.0060, -0.0045, -0.0141], + [ 0.0013, -0.0130, -0.0064, ..., -0.0143, -0.0232, -0.0150]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.mlp.layer2.weight': tensor([[-0.0043, -0.0125, -0.0120, ..., -0.0203, 0.0067, -0.0003], + [-0.0059, 0.0171, 0.0068, ..., 0.0078, 0.0133, 0.0211], + [-0.0038, 0.0016, 0.0134, ..., -0.0035, 0.0135, 0.0177], + ..., + [-0.0152, 0.0171, 0.0059, ..., -0.0117, -0.0147, 0.0052], + [ 0.0100, -0.0058, 0.0141, ..., 0.0024, 0.0131, 0.0106], + [-0.0130, -0.0079, 0.0022, ..., 0.0167, 0.0105, 0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0444, -0.0437, -0.0427, ..., 0.0010, 0.0008, 0.0009], + [-0.0413, -0.0403, -0.0393, ..., 0.0027, 0.0028, 0.0027], + [-0.0396, -0.0386, -0.0374, ..., 0.0022, 0.0023, 0.0023], + ..., + [ 0.0498, 0.0481, 0.0469, ..., -0.0036, -0.0037, -0.0037], + [ 0.0398, 0.0386, 0.0371, ..., -0.0031, -0.0031, -0.0030], + [-0.0396, -0.0386, -0.0378, ..., 0.0023, 0.0024, 0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0062, 0.0070, 0.0070, ..., 0.0150, 0.0175, 0.0078], + [-0.0017, -0.0014, -0.0012, ..., -0.0032, -0.0031, -0.0013], + [ 0.0013, -0.0071, -0.0061, ..., 0.0034, 0.0036, -0.0070], + ..., + [-0.0325, -0.0776, -0.0645, ..., 0.0221, 0.0236, -0.0718], + [-0.0105, -0.0099, -0.0103, ..., 0.0039, 0.0040, -0.0096], + [ 0.0082, 0.0050, 0.0067, ..., 0.0162, 0.0187, 0.0062]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_cross_attn.1.weight': tensor([[-8.2520e-02, -8.0566e-02, -7.7637e-02, ..., -1.0586e-04, + -1.2064e-04, -1.1063e-04], + [ 3.3691e-02, 3.2471e-02, 3.1982e-02, ..., 1.9222e-06, + 2.7180e-05, 1.4877e-04], + [ 3.0884e-02, 2.9785e-02, 2.8809e-02, ..., 2.7657e-04, + 3.0136e-04, 2.0695e-04], + ..., + [ 5.0293e-02, 4.8828e-02, 4.8340e-02, ..., -1.7929e-04, + -1.7262e-04, -1.0538e-04], + [ 5.4932e-02, 5.3467e-02, 5.2002e-02, ..., -1.6785e-04, + -1.0729e-04, -1.5450e-04], + [-2.5635e-02, -2.4902e-02, -2.4292e-02, ..., -6.2943e-05, + 9.0122e-05, 1.1253e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0039, -0.0059, 0.0014, ..., 0.0275, 0.0325, 0.0023], + [ 0.0033, -0.0044, 0.0059, ..., -0.0010, 0.0016, 0.0042], + [ 0.0018, -0.0019, 0.0026, ..., -0.0084, -0.0086, -0.0021], + ..., + [-0.0198, 0.0078, 0.0280, ..., 0.0178, 0.0287, -0.0089], + [-0.0222, 0.0126, 0.0347, ..., -0.0042, -0.0026, -0.0033], + [ 0.0179, -0.0112, -0.0062, ..., -0.0096, -0.0111, 0.0176]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_mlp.1.weight': tensor([[-0.0771, -0.0752, -0.0737, ..., 0.0002, 0.0003, 0.0003], + [ 0.0282, 0.0273, 0.0267, ..., 0.0003, 0.0004, 0.0004], + [-0.0991, -0.0977, -0.0962, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0723, -0.0708, -0.0698, ..., 0.0002, 0.0003, 0.0002], + [-0.0540, -0.0527, -0.0518, ..., 0.0003, 0.0001, 0.0001], + [ 0.0703, 0.0693, 0.0679, ..., -0.0002, -0.0002, -0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_mlp.2.weight': tensor([[-0.0166, -0.0019, -0.0118, ..., -0.0186, -0.0221, 0.0200], + [-0.0018, -0.0050, 0.0067, ..., 0.0004, 0.0036, -0.0035], + [-0.0019, 0.0020, -0.0027, ..., -0.0019, -0.0019, 0.0019], + ..., + [-0.0486, -0.0170, -0.0557, ..., -0.0356, -0.0126, 0.0272], + [ 0.0071, -0.0099, 0.0264, ..., 0.0056, 0.0048, -0.0047], + [ 0.0138, 0.0110, 0.0107, ..., 0.0124, 0.0065, -0.0101]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.after_proj.weight': tensor([[ 0.0066, -0.0078, -0.0059, ..., -0.0055, 0.0026, -0.0152], + [ 0.0037, -0.0080, 0.0184, ..., -0.0098, -0.0061, -0.0095], + [ 0.0123, 0.0111, 0.0305, ..., 0.0052, 0.0048, -0.0076], + ..., + [ 0.0073, 0.0186, -0.0070, ..., 0.0038, -0.0143, -0.0076], + [-0.0108, -0.0059, -0.0055, ..., 0.0070, -0.0120, 0.0055], + [-0.0166, 0.0165, -0.0107, ..., -0.0049, 0.0009, 0.0073]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.after_proj.bias': tensor([ 0.0026, -0.0143, -0.0110, ..., 0.0014, -0.0015, 0.0062], + dtype=torch.bfloat16)} +[06-26 08:28:43|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:28:43|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 08:28:43|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 3 checkpoint. +[06-26 08:28:43|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:1015:compile_tokenizer_if_enabled] Tokenizer compilation disabled +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:146:__init__] Found 3 hint keys across all samples +[06-26 08:28:43|WARNING|cosmos_transfer2/inference.py:148:__init__] Loading the multicontrol model. Multicontrol inference is not strictly equal to single control +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/config.yaml +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['rainy_background_background'] +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample rainy_background_background +[06-26 08:28:43|DEBUG|cosmos_transfer2/inference.py:186:_generate_sample] InferenceArguments(name='rainy_background_background' prompt_path=None prompt='The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. Beyond the junction the surroundings open onto a dense modern downtown of tall glass-and-steel skyscrapers and a distant city skyline veiled in rain and low cloud. The wet road network with painted lane markings, stop lines and pedestrian crosswalks remains in the foreground, multiple vehicles move through the scene with reflections on the asphalt, and the camera follows the moving ego vehicle smoothly. Photorealistic, atmospheric, high detail.' negative_prompt='The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Empty barren horizon, flat low-rise outskirts, blurry melted skyline, distorted warped buildings.' seed=2025 guidance=3 video_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4') max_frames=None context_frame_index=None image_context_path=None num_conditional_frames=1 resolution='720' sigma_max=None num_video_frames_per_chunk=93 num_steps=35 show_control_condition=False show_input=False keep_input_resolution=True edge=EdgeConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4'), control_weight=1.0, mask_path=None, mask_prompt=None, preset_edge_threshold='medium') depth=None vis=BlurConfig(control_path=None, control_weight=0.6, mask_path=None, mask_prompt=None, preset_blur_strength='medium') seg=SegConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/semseg_front.mp4'), control_weight=0.4, mask_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4'), mask_prompt=None, control_prompt=None) guided_generation_mask=None guided_generation_step_threshold=25 guided_generation_foreground_labels=None) +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.json +[06-26 08:28:43|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:28:43|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None, 'vis': None, 'vis_mask': None, 'vis_mask_prompt': None, 'seg': '/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/semseg_front.mp4', 'seg_mask': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4', 'seg_mask_prompt': None} +[06-26 08:28:43|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:472:generate_img2world] Loading input video... +[06-26 08:28:43|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4 +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:492:generate_img2world] Computing prompt text embeddings... +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:501:generate_img2world] Computing negative prompt text embeddings... +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:513:generate_img2world] Processing image context if available... +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:330:read_and_process_image_context] No image context provided. +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:525:generate_img2world] Loading control inputs... +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4 +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:703:read_and_process_control_input] No vis (blur) control input file found, will compute online.. +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/semseg_front.mp4 +[06-26 08:28:44|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:28:45|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4 +[06-26 08:28:45|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:28:45|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 1/2 +[06-26 08:28:52|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 584845 +[06-26 08:28:52|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 0 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:35:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 2/2 +[06-26 08:35:39|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 643774 +[06-26 08:35:39|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 1 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:42:01|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:736:generate_img2world] Average time per chunk: 395.03808387349636 +[06-26 08:42:01|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:42:02|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_edge.mp4 +[06-26 08:42:02|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:42:03|INFO|cosmos_transfer2/inference.py:311:_generate_sample] vis control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_vis.mp4 +[06-26 08:42:04|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:42:04|INFO|cosmos_transfer2/inference.py:311:_generate_sample] seg control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_seg.mp4 +[06-26 08:42:04|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:42:05|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:42:06|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.mp4 diff --git a/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.json b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.json new file mode 100644 index 0000000000000000000000000000000000000000..ee69f2d68e4b4d9c89fb78913c586d6a1cfa96c7 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.json @@ -0,0 +1 @@ +{"name":"rainy_background_background","prompt_path":null,"prompt":"The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. Beyond the junction the surroundings open onto a dense modern downtown of tall glass-and-steel skyscrapers and a distant city skyline veiled in rain and low cloud. The wet road network with painted lane markings, stop lines and pedestrian crosswalks remains in the foreground, multiple vehicles move through the scene with reflections on the asphalt, and the camera follows the moving ego vehicle smoothly. Photorealistic, atmospheric, high detail.","negative_prompt":"The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Empty barren horizon, flat low-rise outskirts, blurry melted skyline, distorted warped buildings.","seed":2025,"guidance":3,"video_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4","max_frames":null,"context_frame_index":null,"image_context_path":null,"num_conditional_frames":1,"resolution":"720","sigma_max":null,"num_video_frames_per_chunk":93,"num_steps":35,"show_control_condition":false,"show_input":false,"keep_input_resolution":true,"edge":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4","control_weight":1.0,"mask_path":null,"mask_prompt":null,"preset_edge_threshold":"medium"},"depth":null,"vis":{"control_path":null,"control_weight":0.6,"mask_path":null,"mask_prompt":null,"preset_blur_strength":"medium"},"seg":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/semseg_front.mp4","control_weight":0.4,"mask_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4","mask_prompt":null,"control_prompt":null},"guided_generation_mask":null,"guided_generation_step_threshold":25,"guided_generation_foreground_labels":null} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.mp4 b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..30b5be17d139a265fba91988c05bf6eab6ad2529 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:81ad28ef3935ca93ef2438905c52729e80e02f2badca4768d9060c9920755635 +size 2700019 diff --git a/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.txt b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.txt new file mode 100644 index 0000000000000000000000000000000000000000..2b49942d17b2595ce6ba274fe488aec4e7e5d98f --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background.txt @@ -0,0 +1 @@ +The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. Beyond the junction the surroundings open onto a dense modern downtown of tall glass-and-steel skyscrapers and a distant city skyline veiled in rain and low cloud. The wet road network with painted lane markings, stop lines and pedestrian crosswalks remains in the foreground, multiple vehicles move through the scene with reflections on the asphalt, and the camera follows the moving ego vehicle smoothly. Photorealistic, atmospheric, high detail. \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_edge.mp4 b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..5cdebe53fd6e5d64099ea77a2b322d3c74554a2d --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d1a2bd621384e9ffdfe7aeacfdc3c277e04360f64aa63573dd720fda2ac14af +size 2253955 diff --git a/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_seg.mp4 b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_seg.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..0c9e871a90f049dd8297c1def9c952664c193d6d --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_seg.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d25de8bd06e2068b6d7f1587361c411644431665a81b18c832ee03f8c8f250c +size 1640719 diff --git a/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_vis.mp4 b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_vis.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..3954839e8e5b25fcae313f021c538aa5f31d799e --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/rainy_background_background_control_vis.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9a9e148f1100e8757fd79f2d91e144f13d1ef9f647745a7550adf9926d0f418 +size 547205 diff --git a/multicontrol_cookbook_grid/rainy/background/background/spec.json b/multicontrol_cookbook_grid/rainy/background/background/spec.json new file mode 100644 index 0000000000000000000000000000000000000000..515b1a5b3509f34c93a49f36e3d47191b6593a9d --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/background/background/spec.json @@ -0,0 +1,21 @@ +{ + "name": "rainy_background_background", + "prompt": "The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. Beyond the junction the surroundings open onto a dense modern downtown of tall glass-and-steel skyscrapers and a distant city skyline veiled in rain and low cloud. The wet road network with painted lane markings, stop lines and pedestrian crosswalks remains in the foreground, multiple vehicles move through the scene with reflections on the asphalt, and the camera follows the moving ego vehicle smoothly. Photorealistic, atmospheric, high detail.", + "video_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4", + "guidance": 3, + "num_steps": 35, + "seed": 2025, + "edge": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4", + "control_weight": 1.0 + }, + "seg": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/semseg_front.mp4", + "control_weight": 0.4, + "mask_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/rainy/background_mask.mp4" + }, + "vis": { + "control_weight": 0.6 + }, + "negative_prompt": "The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Empty barren horizon, flat low-rise outskirts, blurry melted skyline, distorted warped buildings." +} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/car_color/color/config.yaml b/multicontrol_cookbook_grid/rainy/car_color/color/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3c65abbf27d5997e1b6c38f5503cdd3b3148c93c --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/car_color/color/config.yaml @@ -0,0 +1,882 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + keys_to_skip_loading: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'False' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataloaders: + image_data: + ratio: '0' + video_data: + dataloader: + batch_size: '1' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '32' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control_and_image_context + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' + use_cache: 'False' + ratio: '1' + video_data_1: + dataloader: + batch_size: '6' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '32' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control_and_image_context + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + num_video_frames: '1' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' + use_cache: 'False' + ratio: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_control_conditioner +- ema: power +- tokenizer: wan2pt1_tokenizer +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: vid2vid_2B_control + name: vid2vid_2B_control_720p_t24_control_layer4_cr1pt1_embedding_rectified_flow_with_image_context_with_image_data + project: cosmos_transfer2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + base_load_from: null + conditional_frame_timestep: -1.0 + conditional_frames_probs: + 0: 0.4 + 1: 0.4 + 2: 0.2 + conditioner: + _target_: + control_input_depth: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth + output_key: control_input_depth + control_input_depth_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth_mask + output_key: control_input_depth_mask + control_input_edge: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge + output_key: control_input_edge + control_input_edge_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge_mask + output_key: control_input_edge_mask + control_input_inpaint: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint + output_key: control_input_inpaint + control_input_inpaint_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint_mask + output_key: control_input_inpaint_mask + control_input_seg: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg + output_key: control_input_seg + control_input_seg_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg_mask + output_key: control_input_seg_mask + control_input_vis: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis + output_key: control_input_vis + control_input_vis_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis_mask + output_key: control_input_vis_mask + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + reference_image_context: + _target_: + dropout_rate: '0.5' + input_key: + - images + - video + - image_context + num_token: '256' + output_key: null + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + copy_weight_strategy: first_n + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + hint_keys: edge + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + condition_strategy: spaced + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_image_context_dim: '1152' + extra_per_block_abs_pos_emb: 'False' + img_context_deep_proj: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + mlp_ratio: '4.0' + model_channels: '2048' + num_blocks: '28' + num_control_branches: '1' + num_heads: '16' + num_max_modalities: '8' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + separate_embedders: 'False' + share_q_in_i2v_cross_attn: 'False' + spatial_compression_factor: '8' + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_after_proj_for_multi_branch: 'True' + use_crossattn_projection: 'True' + use_cuda_graphs: 'False' + use_input_hint_block: 'False' + use_wan_fp32_strategy: 'True' + vace_block_every_n: '7' + vace_has_mask: 'False' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: uniform + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: false + use_lora: false + use_reference_image: true + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 8 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: '8.63e-05' + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '100000' + f_max: + - '0.5' + f_min: + - '0.2' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'True' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'False' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + frame_loss_log: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '100' + hit_thres: '300' + save_s3: 'True' + save_s3_every_log_n: '10' + load_base_model: + _target_: + config: null + trainer: null + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + compile_config: + recompile_limit: '8' + use_duck_shape: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '100000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/multicontrol_cookbook_grid/rainy/car_color/color/console.log b/multicontrol_cookbook_grid/rainy/car_color/color/console.log new file mode 100644 index 0000000000000000000000000000000000000000..ea87f2f517fcf542efe6d8773d777508db8e23eb --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/car_color/color/console.log @@ -0,0 +1,10 @@ +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:146:__init__] Found 1 hint keys across all samples +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color/config.yaml +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['rainy_car_color_color'] +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample rainy_car_color_color +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.json +[06-26 08:53:31|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None} +[06-26 09:04:02|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color_control_edge.mp4 +[06-26 09:04:02|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 09:04:04|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.mp4 diff --git a/multicontrol_cookbook_grid/rainy/car_color/color/debug.log b/multicontrol_cookbook_grid/rainy/car_color/color/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..c7ae21b4f705e61813ea0590fc90e6410d415427 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/car_color/color/debug.log @@ -0,0 +1,969 @@ +[06-26 08:53:08|DEBUG|packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, training=True, smoke=False, verbose=False, experimental_checkpoints=False) +[06-26 08:53:09|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:53:09|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:53:20|CRITICAL|cosmos_transfer2/_src/interactive/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:53:20|CRITICAL|cosmos_transfer2/_src/predict2/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:53:21|DEBUG|cosmos_transfer2/inference.py:46:__init__] SetupArguments(output_dir=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color') model='edge' checkpoint_path='s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000' experiment='edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5' config_file='cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/config.py' context_parallel_size=1 disable_guardrails=True offload_guardrail_models=False keep_going=True profile=False benchmark=False compile_tokenizer= enable_parallel_tokenizer=False parallel_tokenizer_grid=(-1, -1))(['edge']) +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1... +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2... +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor... +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1_with_control... +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control... +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor_with_control... +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control_and_image_context... +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_with_control_input... +[06-26 08:53:21|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering hdmap_augmentor_for_local_datasets... +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package ('flash_attn_3_nv'_) was not found. +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version 0.21.0, expected at least 0.21.5.dev9. +[06-26 08:53:21|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_large_scale +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_small_overfit +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.experiment_list +[06-26 08:53:21|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.exp_large_scale +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.experiment_list +[06-26 08:53:21|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2.experiments +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot.cosmos_agibot_posttrain_example +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview.cosmos_multiview_example +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview +[06-26 08:53:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview.cosmos_singleview_example +[06-26 08:53:21|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:53:22|INFO|cosmos_transfer2/_src/imaginaire/utils/misc.py:151:set_random_seed] Using random seed 0. +[06-26 08:53:22|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:53:22|WARNING|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[06-26 08:53:22|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[06-26 08:53:22|INFO|cosmos_transfer2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/f176dc95b4a70f53ce01c4b302851595e7322b00/tokenizer.pth +[06-26 08:53:22|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.50 s +[06-26 08:53:25|INFO|cosmos_transfer2/_src/transfer2/networks/siglip2.py:65:get_siglip2_model_processor] +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #4-control_input_edge: + ReMapkey + input key: control_input_edge + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #5-control_input_vis: + ReMapkey + input key: control_input_vis + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #6-control_input_depth: + ReMapkey + input key: control_input_depth + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #7-control_input_seg: + ReMapkey + input key: control_input_seg + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #8-control_input_inpaint: + ReMapkey + input key: control_input_inpaint + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #9-control_input_edge_mask: + ReMapkey + input key: control_input_edge_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge_mask + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #10-control_input_vis_mask: + ReMapkey + input key: control_input_vis_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis_mask + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #11-control_input_depth_mask: + ReMapkey + input key: control_input_depth_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth_mask + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #12-control_input_seg_mask: + ReMapkey + input key: control_input_seg_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg_mask + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #13-control_input_inpaint_mask: + ReMapkey + input key: control_input_inpaint_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint_mask + Dtype: None +[06-26 08:53:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #14-reference_image_context: + SigLip2EmbImgContext + input key: ['images', 'video', 'image_context'] + Param count: 0 + Trainable: None + Dropout rate: 0.5 + Input key: ['images', 'video', 'image_context'] + Output key: ['img_context_emb'] +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.07 s +[06-26 08:53:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.84 s +[06-26 08:53:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 4.21 s +[06-26 08:53:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 4.21 s +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:569:freeze_base_model] +Freezing base model + +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:607:freeze_base_model] ✓ Unfroze img_context_proj +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 0 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 1 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 2 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 3 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 4 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 5 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 6 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 7 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 8 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 9 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 10 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 11 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 12 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 13 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 14 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 15 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 16 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 17 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 18 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 19 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 20 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 21 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 22 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 23 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 24 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 25 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 26 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 27 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 0 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 1 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 2 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 3 +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:779:load_base_model] Done loading the base model checkpoint. +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 0-th block weight to control net's 0-th block +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 1-th block weight to control net's 1-th block +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 2-th block weight to control net's 2-th block +[06-26 08:53:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 3-th block weight to control net's 3-th block +[06-26 08:53:26|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[06-26 08:53:26|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[06-26 08:53:27|INFO|cosmos_transfer2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[06-26 08:53:27|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[06-26 08:53:27|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[06-26 08:53:27|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[06-26 08:53:27|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[06-26 08:53:30|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[06-26 08:53:30|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[06-26 08:53:30|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:76:__init__] MinimalV4LVGControlVaceDiT( + (x_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=72, out_features=2048, bias=False) + ) + ) + (pos_embedder): VideoRopePosition3DEmb() + (t_embedder): Sequential( + (0): Timesteps() + (1): TimestepEmbedding( + (linear_1): Linear(in_features=2048, out_features=2048, bias=False) + (activation): SiLU() + (linear_2): Linear(in_features=2048, out_features=6144, bias=False) + ) + ) + (blocks): ModuleList( + (0-27): 28 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlAwareDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + ) + ) + ) + (final_layer): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): FinalLayer( + (layer_norm): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (linear): Linear(in_features=2048, out_features=64, bias=False) + (adaln_modulation): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=4096, bias=False) + ) + ) + ) + ) + ) + (t_embedding_norm): RMSNorm() + (img_context_proj): Sequential( + (0): Linear(in_features=1152, out_features=2048, bias=True) + (1): GELU(approximate='none') + ) + (crossattn_proj): Sequential( + (0): Linear(in_features=100352, out_features=1024, bias=True) + (1): GELU(approximate='none') + ) + (control_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=520, out_features=2048, bias=False) + ) + ) + (control_blocks): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) +) +[06-26 08:53:30|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 8.13 s +[06-26 08:53:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 08:53:30|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 08:53:31|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:53:31|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +[06-26 08:53:31|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:271:load_model_state_dict_from_checkpoint] Checkpoint weights loaded from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt: None +[06-26 08:53:31|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:273:load_model_state_dict_from_checkpoint] Checkpoint weights loaded successfully (strict=True). +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:1015:compile_tokenizer_if_enabled] Tokenizer compilation disabled +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:146:__init__] Found 1 hint keys across all samples +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color/config.yaml +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['rainy_car_color_color'] +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample rainy_car_color_color +[06-26 08:53:31|DEBUG|cosmos_transfer2/inference.py:186:_generate_sample] InferenceArguments(name='rainy_car_color_color' prompt_path=None prompt='The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. The vehicles moving through the junction are glossy vivid candy-red, their wet paint gleaming and mirroring the grey sky and surrounding lights. A wet road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, low-rise buildings and sidewalks line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, rain-soaked reflections, high detail.' negative_prompt='The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Dull grey washed-out cars, color bleeding onto road or buildings, melted or deformed vehicle bodies.' seed=2025 guidance=3 video_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4') max_frames=None context_frame_index=None image_context_path=None num_conditional_frames=1 resolution='720' sigma_max=None num_video_frames_per_chunk=93 num_steps=35 show_control_condition=False show_input=False keep_input_resolution=True edge=EdgeConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4'), control_weight=1.0, mask_path=None, mask_prompt=None, preset_edge_threshold='medium') depth=None vis=None seg=None guided_generation_mask=None guided_generation_step_threshold=25 guided_generation_foreground_labels=None) +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.json +[06-26 08:53:31|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:53:31|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None} +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:472:generate_img2world] Loading input video... +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4 +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:492:generate_img2world] Computing prompt text embeddings... +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:501:generate_img2world] Computing negative prompt text embeddings... +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:513:generate_img2world] Processing image context if available... +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:330:read_and_process_image_context] No image context provided. +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:525:generate_img2world] Loading control inputs... +[06-26 08:53:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4 +[06-26 08:53:32|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:53:32|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 1/2 +[06-26 08:53:32|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 584845 +[06-26 08:53:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 0 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:53:32|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:255:get_condition_with_negative_prompt] Getting condition with image context +[06-26 08:53:32|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:53:32|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:257:get_condition_with_negative_prompt] Getting uncondition with negative prompt without image context +[06-26 08:53:32|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:53:50|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:58:55|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 2/2 +[06-26 08:58:55|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 643774 +[06-26 08:58:55|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 1 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:58:55|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:255:get_condition_with_negative_prompt] Getting condition with image context +[06-26 08:58:55|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:58:55|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:257:get_condition_with_negative_prompt] Getting uncondition with negative prompt without image context +[06-26 08:58:55|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:58:57|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 09:04:00|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:736:generate_img2world] Average time per chunk: 312.46804093050014 +[06-26 09:04:00|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 09:04:02|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color_control_edge.mp4 +[06-26 09:04:02|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 09:04:03|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 09:04:04|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.mp4 diff --git a/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.json b/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.json new file mode 100644 index 0000000000000000000000000000000000000000..56f3ac30656d256e31ade0c28206166b7ace5e12 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.json @@ -0,0 +1 @@ +{"name":"rainy_car_color_color","prompt_path":null,"prompt":"The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. The vehicles moving through the junction are glossy vivid candy-red, their wet paint gleaming and mirroring the grey sky and surrounding lights. A wet road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, low-rise buildings and sidewalks line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, rain-soaked reflections, high detail.","negative_prompt":"The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Dull grey washed-out cars, color bleeding onto road or buildings, melted or deformed vehicle bodies.","seed":2025,"guidance":3,"video_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4","max_frames":null,"context_frame_index":null,"image_context_path":null,"num_conditional_frames":1,"resolution":"720","sigma_max":null,"num_video_frames_per_chunk":93,"num_steps":35,"show_control_condition":false,"show_input":false,"keep_input_resolution":true,"edge":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4","control_weight":1.0,"mask_path":null,"mask_prompt":null,"preset_edge_threshold":"medium"},"depth":null,"vis":null,"seg":null,"guided_generation_mask":null,"guided_generation_step_threshold":25,"guided_generation_foreground_labels":null} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.mp4 b/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..866d9a9f8ea6c44b4de539aa038b77e6b2eb4a71 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f85880dcbfb464b83384a0d5836a2d2c929294f16f5ab5a431e57a73e4c5fb7b +size 2041880 diff --git a/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.txt b/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.txt new file mode 100644 index 0000000000000000000000000000000000000000..c718e19223d893407bebf3e934465403a9141d49 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color.txt @@ -0,0 +1 @@ +The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. The vehicles moving through the junction are glossy vivid candy-red, their wet paint gleaming and mirroring the grey sky and surrounding lights. A wet road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, low-rise buildings and sidewalks line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, rain-soaked reflections, high detail. \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color_control_edge.mp4 b/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color_control_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..5cdebe53fd6e5d64099ea77a2b322d3c74554a2d --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/car_color/color/rainy_car_color_color_control_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d1a2bd621384e9ffdfe7aeacfdc3c277e04360f64aa63573dd720fda2ac14af +size 2253955 diff --git a/multicontrol_cookbook_grid/rainy/car_color/color/spec.json b/multicontrol_cookbook_grid/rainy/car_color/color/spec.json new file mode 100644 index 0000000000000000000000000000000000000000..40855a516b617909a7f63c71822a25f27fc344d6 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/car_color/color/spec.json @@ -0,0 +1,13 @@ +{ + "name": "rainy_car_color_color", + "prompt": "The video shows an urban intersection captured from a vehicle-mounted camera on a rainy overcast day. The vehicles moving through the junction are glossy vivid candy-red, their wet paint gleaming and mirroring the grey sky and surrounding lights. A wet road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, low-rise buildings and sidewalks line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, rain-soaked reflections, high detail.", + "video_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4", + "guidance": 3, + "num_steps": 35, + "seed": 2025, + "edge": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4", + "control_weight": 1.0 + }, + "negative_prompt": "The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Dull grey washed-out cars, color bleeding onto road or buildings, melted or deformed vehicle bodies." +} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/config.yaml b/multicontrol_cookbook_grid/rainy/weather/lighting/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ddeb6171e5ff728f20e14fb74db291305a4e7054 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/config.yaml @@ -0,0 +1,825 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + keys_to_skip_loading: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'False' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataset: + augmentor_name: video_basic_augmentor_v2_with_control + caption_type: t2w_qwen2p5_7b + control_input_type: edge_vis_depth_segcolor + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_control_conditioner +- ema: power +- tokenizer: wan2pt1_tokenizer +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: vid2vid_2B_control + name: vid2vid_2B_control_720p_t24_control_layer4_cr1pt1_embedding_rectified_flow + project: cosmos_transfer2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + base_load_from: null + conditional_frame_timestep: -1.0 + conditional_frames_probs: + 0: 0.4 + 1: 0.4 + 2: 0.2 + conditioner: + _target_: + control_input_depth: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth + output_key: control_input_depth + control_input_depth_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth_mask + output_key: control_input_depth_mask + control_input_edge: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge + output_key: control_input_edge + control_input_edge_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge_mask + output_key: control_input_edge_mask + control_input_inpaint: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint + output_key: control_input_inpaint + control_input_inpaint_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint_mask + output_key: control_input_inpaint_mask + control_input_seg: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg + output_key: control_input_seg + control_input_seg_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg_mask + output_key: control_input_seg_mask + control_input_vis: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis + output_key: control_input_vis + control_input_vis_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis_mask + output_key: control_input_vis_mask + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + copy_weight_strategy: first_n + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + hint_keys: edge_vis_depth_seg + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + condition_strategy: spaced + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_per_block_abs_pos_emb: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + mlp_ratio: '4.0' + model_channels: '2048' + num_blocks: '28' + num_control_branches: '4' + num_heads: '16' + num_max_modalities: '8' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + separate_embedders: 'False' + spatial_compression_factor: '8' + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_after_proj_for_multi_branch: 'False' + use_crossattn_projection: 'True' + use_cuda_graphs: 'False' + use_input_hint_block: 'False' + use_wan_fp32_strategy: 'True' + vace_block_every_n: '7' + vace_has_mask: 'False' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: reweighting + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: false + use_lora: false + use_reference_image: false + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 8 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: '8.63e-05' + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '100000' + f_max: + - '0.5' + f_min: + - '0.2' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'True' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'False' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + frame_loss_log: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '100' + hit_thres: '300' + save_s3: 'True' + save_s3_every_log_n: '10' + load_base_model: + _target_: + config: null + trainer: null + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + compile_config: + recompile_limit: '8' + use_duck_shape: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '100000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/console.log b/multicontrol_cookbook_grid/rainy/weather/lighting/console.log new file mode 100644 index 0000000000000000000000000000000000000000..84e99c4b82241e53f7b8bcc83081c91716986e59 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/console.log @@ -0,0 +1,12 @@ +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:146:__init__] Found 2 hint keys across all samples +[06-26 08:14:40|WARNING|cosmos_transfer2/inference.py:148:__init__] Loading the multicontrol model. Multicontrol inference is not strictly equal to single control +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/config.yaml +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['rainy_weather_lighting'] +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample rainy_weather_lighting +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.json +[06-26 08:14:40|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None, 'vis': None, 'vis_mask': None, 'vis_mask_prompt': None} +[06-26 08:26:40|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_edge.mp4 +[06-26 08:26:41|INFO|cosmos_transfer2/inference.py:311:_generate_sample] vis control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_vis.mp4 +[06-26 08:26:41|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:26:43|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.mp4 diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/debug.log b/multicontrol_cookbook_grid/rainy/weather/lighting/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..9ff9380c77e229aeb1f999983b9392e3df636c7a --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/debug.log @@ -0,0 +1,33578 @@ +[06-26 08:14:17|DEBUG|packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, training=True, smoke=False, verbose=False, experimental_checkpoints=False) +[06-26 08:14:18|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:14:18|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:14:29|CRITICAL|cosmos_transfer2/_src/interactive/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:14:29|CRITICAL|cosmos_transfer2/_src/predict2/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:14:30|DEBUG|cosmos_transfer2/inference.py:46:__init__] SetupArguments(output_dir=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting') model='edge' checkpoint_path='s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000' experiment='edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5' config_file='cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/config.py' context_parallel_size=1 disable_guardrails=True offload_guardrail_models=False keep_going=True profile=False benchmark=False compile_tokenizer= enable_parallel_tokenizer=False parallel_tokenizer_grid=(-1, -1))(['edge', 'vis']) +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1... +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2... +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor... +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1_with_control... +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control... +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor_with_control... +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control_and_image_context... +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_with_control_input... +[06-26 08:14:30|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering hdmap_augmentor_for_local_datasets... +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package ('flash_attn_3_nv'_) was not found. +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version 0.21.0, expected at least 0.21.5.dev9. +[06-26 08:14:30|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_large_scale +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_small_overfit +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.experiment_list +[06-26 08:14:30|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.exp_large_scale +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.experiment_list +[06-26 08:14:30|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2.experiments +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot.cosmos_agibot_posttrain_example +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview.cosmos_multiview_example +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview +[06-26 08:14:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview.cosmos_singleview_example +[06-26 08:14:30|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:14:30|INFO|cosmos_transfer2/_src/imaginaire/utils/misc.py:151:set_random_seed] Using random seed 0. +[06-26 08:14:30|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:14:30|WARNING|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/f176dc95b4a70f53ce01c4b302851595e7322b00/tokenizer.pth +[06-26 08:14:31|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.50 s +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #4-control_input_edge: + ReMapkey + input key: control_input_edge + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #5-control_input_vis: + ReMapkey + input key: control_input_vis + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #6-control_input_depth: + ReMapkey + input key: control_input_depth + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #7-control_input_seg: + ReMapkey + input key: control_input_seg + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #8-control_input_inpaint: + ReMapkey + input key: control_input_inpaint + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #9-control_input_edge_mask: + ReMapkey + input key: control_input_edge_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge_mask + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #10-control_input_vis_mask: + ReMapkey + input key: control_input_vis_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis_mask + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #11-control_input_depth_mask: + ReMapkey + input key: control_input_depth_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth_mask + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #12-control_input_seg_mask: + ReMapkey + input key: control_input_seg_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg_mask + Dtype: None +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #13-control_input_inpaint_mask: + ReMapkey + input key: control_input_inpaint_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint_mask + Dtype: None +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:14:31|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:14:31|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.09 s +[06-26 08:14:32|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.81 s +[06-26 08:14:32|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 0.81 s +[06-26 08:14:32|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 0.81 s +[06-26 08:14:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:569:freeze_base_model] +Freezing base model + +[06-26 08:14:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:779:load_base_model] Done loading the base model checkpoint. +[06-26 08:14:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 0-th block weight to control net's 0-th block +[06-26 08:14:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 1-th block weight to control net's 1-th block +[06-26 08:14:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 2-th block weight to control net's 2-th block +[06-26 08:14:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 3-th block weight to control net's 3-th block +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[06-26 08:14:32|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[06-26 08:14:32|INFO|cosmos_transfer2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[06-26 08:14:32|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[06-26 08:14:32|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[06-26 08:14:32|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[06-26 08:14:35|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[06-26 08:14:35|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[06-26 08:14:35|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:76:__init__] MinimalV4LVGControlVaceDiT( + (x_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=72, out_features=2048, bias=False) + ) + ) + (pos_embedder): VideoRopePosition3DEmb() + (t_embedder): Sequential( + (0): Timesteps() + (1): TimestepEmbedding( + (linear_1): Linear(in_features=2048, out_features=2048, bias=False) + (activation): SiLU() + (linear_2): Linear(in_features=2048, out_features=6144, bias=False) + ) + ) + (blocks): ModuleList( + (0-27): 28 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlAwareDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + ) + ) + ) + (final_layer): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): FinalLayer( + (layer_norm): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (linear): Linear(in_features=2048, out_features=64, bias=False) + (adaln_modulation): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=4096, bias=False) + ) + ) + ) + ) + ) + ) + ) + ) + (t_embedding_norm): RMSNorm() + (crossattn_proj): Sequential( + (0): Linear(in_features=100352, out_features=1024, bias=True) + (1): GELU(approximate='none') + ) + (control_embedder): ModuleList( + (0-3): 4 x PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=520, out_features=2048, bias=False) + ) + ) + ) + (control_blocks_0): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_1): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_2): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_3): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) +) +[06-26 08:14:35|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 4.79 s +[06-26 08:14:35|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 08:14:35|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 08:14:36|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:36|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'control_blocks.0.self_attn.q_proj.weight', 'control_blocks.0.self_attn.q_norm.weight', 'control_blocks.0.self_attn.k_proj.weight', 'control_blocks.0.self_attn.k_norm.weight', 'control_blocks.0.self_attn.v_proj.weight', 'control_blocks.0.self_attn.output_proj.weight', 'control_blocks.0.cross_attn.q_proj.weight', 'control_blocks.0.cross_attn.q_norm.weight', 'control_blocks.0.cross_attn.k_proj.weight', 'control_blocks.0.cross_attn.k_norm.weight', 'control_blocks.0.cross_attn.v_proj.weight', 'control_blocks.0.cross_attn.output_proj.weight', 'control_blocks.0.cross_attn.k_img.weight', 'control_blocks.0.cross_attn.v_img.weight', 'control_blocks.0.cross_attn.q_img.weight', 'control_blocks.0.cross_attn.q_img_norm.weight', 'control_blocks.0.cross_attn.k_img_norm.weight', 'control_blocks.0.mlp.layer1.weight', 'control_blocks.0.mlp.layer2.weight', 'control_blocks.0.adaln_modulation_self_attn.1.weight', 'control_blocks.0.adaln_modulation_self_attn.2.weight', 'control_blocks.0.adaln_modulation_cross_attn.1.weight', 'control_blocks.0.adaln_modulation_cross_attn.2.weight', 'control_blocks.0.adaln_modulation_mlp.1.weight', 'control_blocks.0.adaln_modulation_mlp.2.weight', 'control_blocks.0.before_proj.weight', 'control_blocks.0.before_proj.bias', 'control_blocks.0.after_proj.weight', 'control_blocks.0.after_proj.bias', 'control_blocks.1.self_attn.q_proj.weight', 'control_blocks.1.self_attn.q_norm.weight', 'control_blocks.1.self_attn.k_proj.weight', 'control_blocks.1.self_attn.k_norm.weight', 'control_blocks.1.self_attn.v_proj.weight', 'control_blocks.1.self_attn.output_proj.weight', 'control_blocks.1.cross_attn.q_proj.weight', 'control_blocks.1.cross_attn.q_norm.weight', 'control_blocks.1.cross_attn.k_proj.weight', 'control_blocks.1.cross_attn.k_norm.weight', 'control_blocks.1.cross_attn.v_proj.weight', 'control_blocks.1.cross_attn.output_proj.weight', 'control_blocks.1.cross_attn.k_img.weight', 'control_blocks.1.cross_attn.v_img.weight', 'control_blocks.1.cross_attn.q_img.weight', 'control_blocks.1.cross_attn.q_img_norm.weight', 'control_blocks.1.cross_attn.k_img_norm.weight', 'control_blocks.1.mlp.layer1.weight', 'control_blocks.1.mlp.layer2.weight', 'control_blocks.1.adaln_modulation_self_attn.1.weight', 'control_blocks.1.adaln_modulation_self_attn.2.weight', 'control_blocks.1.adaln_modulation_cross_attn.1.weight', 'control_blocks.1.adaln_modulation_cross_attn.2.weight', 'control_blocks.1.adaln_modulation_mlp.1.weight', 'control_blocks.1.adaln_modulation_mlp.2.weight', 'control_blocks.1.after_proj.weight', 'control_blocks.1.after_proj.bias', 'control_blocks.2.self_attn.q_proj.weight', 'control_blocks.2.self_attn.q_norm.weight', 'control_blocks.2.self_attn.k_proj.weight', 'control_blocks.2.self_attn.k_norm.weight', 'control_blocks.2.self_attn.v_proj.weight', 'control_blocks.2.self_attn.output_proj.weight', 'control_blocks.2.cross_attn.q_proj.weight', 'control_blocks.2.cross_attn.q_norm.weight', 'control_blocks.2.cross_attn.k_proj.weight', 'control_blocks.2.cross_attn.k_norm.weight', 'control_blocks.2.cross_attn.v_proj.weight', 'control_blocks.2.cross_attn.output_proj.weight', 'control_blocks.2.cross_attn.k_img.weight', 'control_blocks.2.cross_attn.v_img.weight', 'control_blocks.2.cross_attn.q_img.weight', 'control_blocks.2.cross_attn.q_img_norm.weight', 'control_blocks.2.cross_attn.k_img_norm.weight', 'control_blocks.2.mlp.layer1.weight', 'control_blocks.2.mlp.layer2.weight', 'control_blocks.2.adaln_modulation_self_attn.1.weight', 'control_blocks.2.adaln_modulation_self_attn.2.weight', 'control_blocks.2.adaln_modulation_cross_attn.1.weight', 'control_blocks.2.adaln_modulation_cross_attn.2.weight', 'control_blocks.2.adaln_modulation_mlp.1.weight', 'control_blocks.2.adaln_modulation_mlp.2.weight', 'control_blocks.2.after_proj.weight', 'control_blocks.2.after_proj.bias', 'control_blocks.3.self_attn.q_proj.weight', 'control_blocks.3.self_attn.q_norm.weight', 'control_blocks.3.self_attn.k_proj.weight', 'control_blocks.3.self_attn.k_norm.weight', 'control_blocks.3.self_attn.v_proj.weight', 'control_blocks.3.self_attn.output_proj.weight', 'control_blocks.3.cross_attn.q_proj.weight', 'control_blocks.3.cross_attn.q_norm.weight', 'control_blocks.3.cross_attn.k_proj.weight', 'control_blocks.3.cross_attn.k_norm.weight', 'control_blocks.3.cross_attn.v_proj.weight', 'control_blocks.3.cross_attn.output_proj.weight', 'control_blocks.3.cross_attn.k_img.weight', 'control_blocks.3.cross_attn.v_img.weight', 'control_blocks.3.cross_attn.q_img.weight', 'control_blocks.3.cross_attn.q_img_norm.weight', 'control_blocks.3.cross_attn.k_img_norm.weight', 'control_blocks.3.mlp.layer1.weight', 'control_blocks.3.mlp.layer2.weight', 'control_blocks.3.adaln_modulation_self_attn.1.weight', 'control_blocks.3.adaln_modulation_self_attn.2.weight', 'control_blocks.3.adaln_modulation_cross_attn.1.weight', 'control_blocks.3.adaln_modulation_cross_attn.2.weight', 'control_blocks.3.adaln_modulation_mlp.1.weight', 'control_blocks.3.adaln_modulation_mlp.2.weight', 'control_blocks.3.after_proj.weight', 'control_blocks.3.after_proj.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_embedder.proj.1.weight'], incorrect_shapes=[]) +[06-26 08:14:36|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:271:load_model_state_dict_from_checkpoint] Checkpoint weights loaded from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt: None +[06-26 08:14:36|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:273:load_model_state_dict_from_checkpoint] Checkpoint weights loaded successfully (strict=True). +[06-26 08:14:36|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 08:14:36|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 0 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 08:14:37|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0148, 0.0025, 0.0065], + [-0.0004, 0.0057, -0.0327, ..., -0.0078, 0.0021, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0136, -0.0035, 0.0283], + ..., + [-0.0028, 0.0125, 0.0248, ..., 0.0437, -0.0069, -0.0192], + [ 0.0277, 0.0146, 0.0031, ..., -0.0028, -0.0090, 0.0146], + [ 0.0437, 0.0366, 0.0082, ..., 0.0029, -0.0126, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0044, -0.0310, 0.0300, ..., -0.0064, -0.0045, -0.0339], + [-0.0103, -0.0047, 0.0114, ..., 0.0052, 0.0092, 0.0520], + [-0.0061, 0.0131, 0.0398, ..., -0.0096, -0.0179, -0.0056], + ..., + [-0.0048, 0.0175, -0.0101, ..., 0.0225, 0.0043, 0.0042], + [-0.0209, -0.0081, 0.0089, ..., -0.0308, 0.0108, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0055, 0.0354, -0.0082, ..., 0.0120, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0013, ..., 0.0376, 0.0532, 0.0036], + [ 0.0175, -0.0347, 0.0013, ..., -0.0066, 0.0053, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0154, 0.0262, -0.0112], + [-0.0006, 0.0244, 0.0356, ..., -0.0036, 0.0170, 0.0118], + [ 0.0486, 0.0175, 0.0042, ..., 0.0036, -0.0026, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0114, ..., -0.0070, -0.0337, 0.0077], + [ 0.0491, 0.0413, 0.0251, ..., -0.0030, 0.0393, 0.0042], + [-0.0162, -0.0019, 0.0140, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0038, 0.0261, 0.0039, ..., -0.0040, 0.0347, 0.0009], + [-0.0254, 0.0063, 0.0019, ..., 0.0093, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0121, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0035, -0.0270, -0.0166, ..., 0.0115, -0.0298, 0.0532], + [ 0.0017, 0.0040, -0.0131, ..., -0.0161, 0.0146, 0.0308], + [ 0.0146, 0.0087, 0.0143, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0064, -0.0630, -0.0048, ..., 0.0074, -0.0183, 0.0217], + [ 0.0023, -0.0035, 0.0258, ..., -0.0486, -0.0349, -0.0059], + [ 0.0182, -0.0076, 0.0197, ..., 0.0142, -0.0226, -0.0212]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0051, -0.0012, -0.0011, ..., -0.0005, -0.0018, -0.0120], + [-0.0153, 0.0255, 0.0432, ..., -0.0045, 0.0049, -0.0069], + [-0.0154, -0.0206, -0.0245, ..., -0.0119, -0.0325, -0.0013], + ..., + [-0.0034, 0.0039, 0.0417, ..., 0.0280, 0.0036, -0.0047], + [-0.0054, 0.0041, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0055, -0.0024, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 1.5503e-02, -6.5613e-03, -4.8096e-02, ..., 4.4922e-02, + 2.1240e-02, 1.2756e-02], + [ 1.5381e-02, -2.2705e-02, 2.1973e-02, ..., -2.2095e-02, + 3.5858e-03, 6.9885e-03], + [-1.5259e-02, -1.0071e-02, 4.2969e-02, ..., -9.5825e-03, + 3.2715e-02, -1.1169e-02], + ..., + [ 3.1006e-02, -3.7598e-02, -4.3457e-02, ..., -2.2461e-02, + 2.5513e-02, 1.3428e-02], + [-3.8862e-05, 1.6602e-02, -1.6174e-03, ..., -4.0283e-03, + 6.7139e-03, -9.6436e-03], + [-1.0864e-02, -3.4637e-03, 1.1780e-02, ..., 4.6997e-03, + -3.7994e-03, -7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[ 0.0049, -0.0386, -0.0112, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0079, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0127, 0.0032, -0.0114, ..., 0.0214, 0.0110, -0.0045], + ..., + [-0.0155, -0.0172, -0.0136, ..., 0.0095, 0.0098, 0.0184], + [-0.0151, -0.0356, 0.0200, ..., -0.0060, -0.0112, 0.0332], + [-0.0396, -0.0120, -0.0089, ..., -0.0284, -0.0140, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 2.7710e-02, 3.1250e-02, 1.3306e-02, ..., -1.9531e-02, + 1.1414e-02, -1.4954e-02], + [ 1.9165e-02, 9.2773e-03, -7.5378e-03, ..., -1.7853e-03, + -1.2695e-02, -1.9775e-02], + [-2.5024e-02, 1.1108e-02, -1.3733e-02, ..., 1.0132e-02, + 1.7212e-02, 7.0496e-03], + ..., + [ 4.6631e-02, 6.4392e-03, -1.5869e-03, ..., -4.0283e-03, + 2.3926e-02, 2.5330e-03], + [ 5.4688e-02, -1.8677e-02, -1.1108e-02, ..., -3.7537e-03, + 3.1128e-02, -5.4626e-03], + [ 8.7738e-05, 1.0742e-02, -9.6436e-03, ..., 2.0020e-02, + 1.6968e-02, 1.4038e-02]], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[-5.6028e-05, 2.3651e-03, 2.8564e-02, ..., -1.0376e-02, + 4.5471e-03, -1.8188e-02], + [ 6.2866e-03, 1.3428e-02, -3.1494e-02, ..., -4.5654e-02, + -5.0049e-03, 1.0376e-02], + [ 4.3945e-03, 2.6123e-02, 2.3926e-02, ..., 5.0049e-03, + 6.6833e-03, 4.6143e-02], + ..., + [ 1.6968e-02, -2.5391e-02, -4.3945e-02, ..., 1.5259e-02, + 4.9133e-03, -7.4768e-03], + [ 5.7678e-03, -3.7109e-02, 3.5645e-02, ..., -8.3008e-03, + 4.1016e-02, -1.7822e-02], + [ 1.7822e-02, -1.4648e-02, 5.6152e-02, ..., -2.9053e-02, + -2.6733e-02, 2.9663e-02]], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0067, 0.0208, 0.0278, ..., 0.0133, 0.0156, 0.0074], + [-0.0063, 0.0040, 0.0014, ..., 0.0135, 0.0040, 0.0045], + [ 0.0058, 0.0041, 0.0243, ..., -0.0123, -0.0049, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0064, ..., 0.0250, -0.0107, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., 0.0048, -0.0295, 0.0400], + [-0.0116, 0.0138, 0.0305, ..., -0.0103, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0054, 0.0381, -0.0239, ..., 0.0219, 0.0483, 0.0339], + [-0.0143, -0.0033, 0.0288, ..., 0.0135, 0.0583, 0.0049], + [-0.0280, 0.0154, -0.0386, ..., -0.0444, -0.0092, -0.0026], + ..., + [-0.0038, 0.0157, 0.0217, ..., 0.0048, 0.0114, 0.0154], + [ 0.0081, 0.0195, -0.0283, ..., -0.0060, -0.0021, -0.0342], + [-0.0166, -0.0195, 0.0027, ..., -0.0066, 0.0165, 0.0130]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0138, -0.0146, 0.0459, ..., -0.0028, -0.0320, 0.0277], + [-0.0026, 0.0068, 0.0270, ..., -0.0220, -0.0183, -0.0271], + [ 0.0221, 0.0219, -0.0201, ..., 0.0060, 0.0064, -0.0031], + ..., + [-0.0258, 0.0016, 0.0215, ..., 0.0038, 0.0237, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0055], + [-0.0006, -0.0080, -0.0120, ..., -0.0334, 0.0047, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0147, -0.0454, ..., -0.0132, -0.0125, 0.0400], + [ 0.0295, -0.0187, -0.0110, ..., -0.0283, 0.0047, -0.0186], + [-0.0007, 0.0547, 0.0024, ..., 0.0116, 0.0400, -0.0214], + ..., + [ 0.0280, -0.0058, -0.0266, ..., -0.0084, -0.0287, 0.0150], + [ 0.0090, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0095, 0.0102, ..., -0.0070, 0.0703, -0.0043]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0205, 0.0087, -0.0315, ..., 0.0081, 0.0254, 0.0359], + [ 0.0097, -0.0032, 0.0038, ..., -0.0120, -0.0233, -0.0048], + [ 0.0051, 0.0040, -0.0121, ..., -0.0117, -0.0070, 0.0086], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0040], + [ 0.0339, -0.0064, 0.0223, ..., 0.0410, -0.0017, 0.0136], + [-0.0046, -0.0439, -0.0486, ..., 0.0145, -0.0130, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0046, 0.0139, -0.0030, ..., -0.0081, -0.0033, -0.0229], + [-0.0289, 0.0231, 0.0077, ..., -0.0135, -0.0024, 0.0222], + [ 0.0083, 0.0295, -0.0189, ..., -0.0168, 0.0311, -0.0261], + ..., + [ 0.0233, -0.0045, -0.0036, ..., 0.0032, 0.0278, 0.0049], + [-0.0106, -0.0038, 0.0140, ..., -0.0175, 0.0444, 0.0425], + [ 0.0069, -0.0078, 0.0102, ..., 0.0066, -0.0344, 0.0229]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[-0.0050, 0.0128, 0.0190, ..., -0.0093, -0.0198, 0.0149], + [-0.0581, 0.0294, -0.0047, ..., -0.0240, 0.0043, 0.0300], + [ 0.0085, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0035, -0.0289, 0.0537, ..., 0.0058, -0.0074, -0.0099], + [-0.0029, -0.0069, 0.0210, ..., -0.0021, 0.0091, 0.0186], + [-0.0086, -0.0349, -0.0173, ..., -0.0400, 0.0052, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0177, -0.0261, -0.0112, ..., 0.0110, -0.0130, -0.0405], + [ 0.0058, -0.0086, 0.0266, ..., -0.0400, 0.0073, 0.0391], + [-0.0013, 0.0359, 0.0284, ..., 0.0286, -0.0234, 0.0217], + ..., + [ 0.0017, 0.0123, 0.0132, ..., -0.0138, 0.0032, 0.0096], + [ 0.0299, -0.0167, -0.0022, ..., -0.0063, 0.0043, -0.0206], + [ 0.0074, -0.0076, -0.0043, ..., -0.0047, -0.0067, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, -0.0032, 0.0121, ..., -0.0075, -0.0208, -0.0058], + [-0.0092, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0229], + [ 0.0132, -0.0278, 0.0115, ..., 0.0396, -0.0082, -0.0074], + ..., + [-0.0015, 0.0018, -0.0309, ..., 0.0344, 0.0073, -0.0126], + [-0.0236, -0.0102, -0.0056, ..., 0.0332, -0.0117, -0.0171], + [ 0.0231, 0.0405, 0.0103, ..., -0.0278, -0.0139, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[-0.0046, 0.0166, 0.0078, ..., 0.0273, -0.0352, 0.0153], + [-0.0510, -0.0070, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0077, ..., -0.0388, -0.0047, 0.0398], + ..., + [-0.0160, -0.0195, -0.0051, ..., -0.0040, -0.0055, -0.0035], + [-0.0187, -0.0040, 0.0097, ..., 0.0287, -0.0122, -0.0099], + [-0.0181, -0.0013, 0.0247, ..., -0.0145, 0.0072, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0078, 0.0092, 0.0140, ..., -0.0121, 0.0045, -0.0253], + [-0.0164, 0.0330, -0.0171, ..., -0.0033, -0.0403, -0.0111], + [ 0.0085, 0.0347, 0.0275, ..., 0.0151, 0.0069, 0.0080], + ..., + [ 0.0275, -0.0168, -0.0106, ..., -0.0099, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0056, ..., -0.0023, -0.0022, -0.0243], + [-0.0060, 0.0581, 0.0061, ..., 0.0148, 0.0239, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0121, -0.0410, 0.0128, ..., -0.0254, -0.0036, -0.0173], + [ 0.0476, 0.0315, 0.0141, ..., -0.0154, -0.0073, 0.0048], + [-0.0092, -0.0270, 0.0092, ..., 0.0332, 0.0150, -0.0073], + ..., + [-0.0076, 0.0039, 0.0162, ..., -0.0177, 0.0171, -0.0133], + [ 0.0044, -0.0187, -0.0051, ..., 0.0248, -0.0145, -0.0073], + [-0.0143, 0.0034, 0.0149, ..., -0.0047, -0.0006, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0049, 0.0097, ..., 0.0255, -0.0253, -0.0136], + [-0.0038, -0.0141, -0.0160, ..., -0.0109, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0117, 0.0056, 0.0322, ..., -0.0183, -0.0041, -0.0062], + [ 0.0045, -0.0054, 0.0164, ..., 0.0036, 0.0255, 0.0024], + [ 0.0270, 0.0084, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0077, ..., -0.0532, 0.0106, 0.0036], + [-0.0240, -0.0215, -0.0104, ..., -0.0063, 0.0097, -0.0168], + [ 0.0075, 0.0064, -0.0055, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0059, 0.0172, -0.0106, ..., -0.0300, -0.0005, 0.0435], + [ 0.0053, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0061], + [-0.0167, 0.0128, 0.0101, ..., -0.0247, 0.0119, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0065, -0.0079, 0.0317, ..., 0.0200, -0.0105, -0.0376], + [ 0.0356, 0.0378, 0.0129, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., 0.0076, 0.0065, -0.0322], + ..., + [ 0.0065, 0.0107, -0.0029, ..., 0.0366, -0.0206, 0.0374], + [-0.0034, -0.0079, -0.0044, ..., -0.0095, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0048, ..., -0.0166, -0.0042, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0078, -0.0096, -0.0344, ..., -0.0003, 0.0347, -0.0181], + [ 0.0267, -0.0366, -0.0089, ..., 0.0217, 0.0349, -0.0070], + [-0.0098, -0.0286, 0.0062, ..., 0.0060, -0.0282, 0.0047], + ..., + [-0.0124, -0.0039, 0.0065, ..., -0.0084, 0.0165, 0.0074], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0134], + [-0.0140, 0.0049, -0.0170, ..., -0.0126, 0.0060, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0033, 0.0051, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0120, 0.0374, ..., -0.0153, -0.0188, -0.0137], + ..., + [ 0.0262, 0.0052, -0.0065, ..., 0.0007, 0.0038, -0.0165], + [-0.0222, -0.0155, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0125, -0.0148, 0.0038, ..., 0.0018, -0.0234, 0.0017]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0102], + [ 0.0249, 0.0205, 0.0054, ..., -0.0046, -0.0093, -0.0106], + [-0.0220, -0.0004, -0.0037, ..., 0.0306, 0.0381, -0.0044], + ..., + [ 0.0200, -0.0050, 0.0170, ..., 0.0116, 0.0320, -0.0157], + [ 0.0349, 0.0119, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0049, 0.0256, 0.0045, ..., 0.0205, 0.0002, -0.0105]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0227, 0.0065, 0.0199, ..., -0.0057, 0.0276, -0.0312], + [ 0.0178, -0.0059, 0.0103, ..., 0.0112, -0.0398, 0.0293], + [ 0.0430, -0.0130, 0.0072, ..., 0.0040, -0.0069, -0.0029], + ..., + [ 0.0192, -0.0112, -0.0112, ..., 0.0027, 0.0330, 0.0137], + [-0.0125, -0.0542, -0.0205, ..., 0.0044, -0.0159, 0.0320], + [ 0.0303, 0.0047, -0.0425, ..., 0.0032, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0060, -0.0315, ..., 0.0525, 0.0041, -0.0349], + [-0.0036, -0.0026, -0.0156, ..., 0.0048, -0.0240, 0.0305], + [-0.0256, 0.0064, 0.0065, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0065, 0.0231, 0.0209, ..., 0.0076, -0.0041, 0.0461], + [-0.0063, -0.0079, -0.0019, ..., -0.0217, 0.0347, -0.0040], + [ 0.0050, 0.0234, 0.0031, ..., 0.0261, -0.0052, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0052, 0.0046], + [-0.0183, -0.0211, -0.0052, ..., -0.0149, 0.0139, -0.0162], + [-0.0078, -0.0160, 0.0079, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0096, 0.0073, ..., 0.0171, 0.0036, -0.0033], + [-0.0144, 0.0260, -0.0087, ..., 0.0041, 0.0106, -0.0166], + [ 0.0098, -0.0140, -0.0014, ..., 0.0057, 0.0054, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0078, 0.0134, -0.0064, ..., -0.0278, 0.0240, -0.0123], + [-0.0069, 0.0188, -0.0160, ..., -0.0317, -0.0064, 0.0181], + [-0.0042, 0.0310, -0.0085, ..., -0.0226, 0.0017, 0.0334], + ..., + [-0.0081, 0.0186, -0.0073, ..., 0.0228, 0.0047, 0.0064], + [ 0.0271, 0.0118, 0.0354, ..., -0.0075, 0.0198, 0.0150], + [ 0.0282, 0.0171, 0.0317, ..., -0.0109, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-1.1475e-02, -2.6489e-02, -1.0986e-02, ..., 4.8523e-03, + 3.7231e-03, 1.3428e-02], + [-3.3203e-02, -4.8523e-03, 3.0640e-02, ..., 3.4668e-02, + -3.6133e-02, 8.4839e-03], + [-5.8651e-05, -1.9043e-02, -3.6774e-03, ..., 1.6602e-02, + -1.3550e-02, 1.3733e-02], + ..., + [-2.0020e-02, -2.9144e-03, 2.4414e-04, ..., -3.3936e-02, + 3.6133e-02, 3.0518e-03], + [ 6.6528e-03, -2.5024e-02, 1.5564e-02, ..., 2.2430e-03, + -1.6235e-02, 2.1973e-02], + [-7.4463e-03, 3.8452e-03, 2.0508e-02, ..., -6.5002e-03, + -8.3008e-03, 4.3457e-02]], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0047, 0.0240, -0.0217, ..., 0.0172, 0.0046, 0.0114], + [-0.0386, 0.0106, 0.0078, ..., -0.0060, 0.0322, 0.0086], + [-0.0427, 0.0011, -0.0045, ..., 0.0076, 0.0112, -0.0137], + ..., + [-0.0087, -0.0037, -0.0264, ..., -0.0586, -0.0630, -0.0131], + [-0.0006, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0107, -0.0610, 0.0162, ..., -0.0291, -0.0166, 0.0029]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0117, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0112], + [ 0.0111, -0.0098, -0.0327, ..., 0.0447, 0.0076, 0.0059], + [ 0.0042, -0.0175, 0.0076, ..., -0.0069, -0.0037, -0.0060], + ..., + [ 0.0295, 0.0042, -0.0102, ..., 0.0099, -0.0160, -0.0286], + [ 0.0082, 0.0253, -0.0177, ..., 0.0095, 0.0064, 0.0078], + [-0.0132, -0.0048, 0.0139, ..., -0.0125, 0.0277, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0129, 0.0040, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0050, 0.0234, 0.0069, ..., 0.0464, 0.0056, 0.0309], + ..., + [-0.0118, -0.0175, 0.0327, ..., 0.0132, 0.0051, 0.0141], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, 0.0040, -0.0444, ..., 0.0060, -0.0283, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0081, -0.0100, 0.0146, ..., -0.0742, -0.0074, 0.0161], + [-0.0064, -0.0132, -0.0071, ..., 0.0053, 0.0042, 0.0238], + [ 0.0189, 0.0221, -0.0087, ..., 0.0033, 0.0408, 0.0064], + ..., + [ 0.0047, 0.0247, 0.0396, ..., 0.0032, -0.0304, -0.0183], + [ 0.0299, -0.0113, -0.0417, ..., -0.0052, -0.0376, 0.0032], + [-0.0210, 0.0168, -0.0039, ..., 0.0052, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0056, 0.0242, -0.0079, ..., -0.0245, 0.0050, -0.0057], + [ 0.0190, -0.0123, 0.0332, ..., -0.0143, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0053, ..., 0.0056, -0.0091, 0.0182], + ..., + [-0.0430, -0.0403, -0.0086, ..., 0.0134, -0.0140, 0.0060], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0089, 0.0378], + [ 0.0364, 0.0486, -0.0120, ..., -0.0060, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0094, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0078, ..., 0.0101, -0.0223, -0.0052], + ..., + [ 0.0192, 0.0292, -0.0022, ..., 0.0176, -0.0197, 0.0076], + [ 0.0265, 0.0064, 0.0057, ..., 0.0286, 0.0046, -0.0267], + [-0.0184, -0.0088, -0.0045, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0083, -0.0159, 0.0082, ..., 0.0198, 0.0044, 0.0277], + [ 0.0172, 0.0037, -0.0053, ..., 0.0054, -0.0044, 0.0074], + [-0.0019, 0.0184, 0.0113, ..., 0.0099, -0.0054, -0.0077], + ..., + [ 0.0061, -0.0111, -0.0028, ..., -0.0095, -0.0038, 0.0123], + [-0.0427, -0.0391, -0.0193, ..., 0.0096, -0.0238, -0.0078], + [ 0.0040, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0217, -0.0226, -0.0349, ..., -0.0079, -0.0161, 0.0127], + [-0.0049, 0.0008, 0.0312, ..., 0.0055, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0076, 0.0004, 0.0278], + ..., + [ 0.0170, -0.0082, 0.0471, ..., 0.0018, 0.0083, -0.0049], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0079, -0.0203], + [ 0.0033, 0.0081, -0.0077, ..., -0.0098, -0.0076, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[-0.0044, -0.0151, 0.0284, ..., -0.0109, 0.0140, -0.0002], + [ 0.0156, 0.0332, 0.0337, ..., 0.0223, -0.0165, -0.0015], + [-0.0071, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0068, 0.0062, 0.0280, ..., -0.0066, -0.0072, 0.0137], + [-0.0073, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0204], + [-0.0206, 0.0159, -0.0091, ..., 0.0045, 0.0167, -0.0086]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0055, ..., -0.0238, -0.0102, 0.0119], + [-0.0250, -0.0162, 0.0019, ..., -0.0084, 0.0035, -0.0369], + [ 0.0004, 0.0141, 0.0061, ..., -0.0452, -0.0056, 0.0121], + ..., + [ 0.0147, -0.0140, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0099, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0141, -0.0048, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[ 0.0013, -0.0327, -0.0190, ..., 0.0053, -0.0018, -0.0209], + [ 0.0713, 0.0150, -0.0026, ..., 0.0073, 0.0186, -0.0029], + [-0.0227, -0.0090, 0.0188, ..., -0.0125, 0.0193, -0.0049], + ..., + [ 0.0117, 0.0464, 0.0079, ..., 0.0248, 0.0168, -0.0038], + [ 0.0381, 0.0045, 0.0067, ..., 0.0048, -0.0110, 0.0035], + [ 0.0292, 0.0055, 0.0209, ..., 0.0044, 0.0234, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0063, 0.0212, ..., 0.0305, 0.0137, 0.0104], + [-0.0072, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0050], + ..., + [ 0.0082, -0.0275, 0.0225, ..., 0.0051, -0.0189, -0.0104], + [-0.0125, -0.0051, -0.0041, ..., 0.0046, -0.0048, 0.0211], + [ 0.0046, -0.0064, -0.0060, ..., -0.0109, 0.0264, 0.0047]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[-0.0002, -0.0076, -0.0109, ..., 0.0186, -0.0216, 0.0062], + [ 0.0045, 0.0123, -0.0231, ..., 0.0125, -0.0114, -0.0183], + [ 0.0107, 0.0103, -0.0038, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0121, 0.0146, ..., 0.0540, 0.0415, 0.0096], + [ 0.0061, 0.0084, -0.0143, ..., 0.0060, -0.0189, -0.0237], + [ 0.0048, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0067]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0019, 0.0359, ..., -0.0083, -0.0123, -0.0050], + [ 0.0190, -0.0044, 0.0298, ..., 0.0127, 0.0042, -0.0200], + [ 0.0114, -0.0112, 0.0177, ..., -0.0248, 0.0229, 0.0025], + ..., + [-0.0082, -0.0049, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0042, -0.0291, ..., 0.0206, 0.0304, -0.0089], + [-0.0081, 0.0245, -0.0066, ..., -0.0413, 0.0216, -0.0117]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0078, -0.0157, 0.0077, ..., -0.0049, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0064], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, 0.0038], + ..., + [-0.0036, 0.0003, 0.0258, ..., -0.0310, 0.0091, -0.0361], + [-0.0078, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [ 0.0067, 0.0157, -0.0070, ..., 0.0040, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0077, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0106, ..., 0.0103, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0041, ..., 0.0129, -0.0046, 0.0085], + ..., + [-0.0225, 0.0337, -0.0085, ..., -0.0177, 0.0542, -0.0249], + [ 0.0013, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0128, 0.0206, ..., 0.0071, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0095, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0052, 0.0063, ..., 0.0039, 0.0113, 0.0598], + [-0.0068, 0.0267, 0.0054, ..., 0.0284, -0.0060, -0.0364], + ..., + [ 0.0044, 0.0040, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0046, -0.0089, ..., 0.0211, 0.0135, -0.0075], + [ 0.0101, 0.0236, -0.0206, ..., -0.0135, -0.0048, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0067, 0.0294, ..., 0.0050, 0.0270, 0.0063], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0065, -0.0168], + [ 0.0201, -0.0120, 0.0126, ..., -0.0381, -0.0247, -0.0076], + ..., + [ 0.0275, -0.0182, -0.0183, ..., 0.0464, 0.0091, -0.0077], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0067, -0.0466], + [ 0.0327, 0.0284, -0.0030, ..., 0.0143, 0.0142, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, -0.0027, -0.0167, ..., -0.0039, 0.0170, 0.0107], + [-0.0002, -0.0002, 0.0295, ..., -0.0172, 0.0048, 0.0135], + [ 0.0193, 0.0236, -0.0123, ..., -0.0007, -0.0369, 0.0239], + ..., + [ 0.0150, -0.0413, -0.0126, ..., -0.0354, 0.0052, -0.0320], + [-0.0265, -0.0078, -0.0327, ..., -0.0070, 0.0077, 0.0074], + [ 0.0299, 0.0356, -0.0192, ..., 0.0059, -0.0052, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0013, 0.0041, 0.0317, ..., 0.0042, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0076, -0.0093], + [-0.0166, 0.0072, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0019, 0.0260, ..., 0.0159, 0.0048, -0.0474], + [-0.0056, 0.0080, 0.0077, ..., 0.0146, 0.0557, 0.0048], + [-0.0203, 0.0273, 0.0047, ..., -0.0330, 0.0065, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0058, ..., -0.0014, -0.0093, -0.0277], + [ 0.0060, -0.0078, -0.0062, ..., 0.0036, 0.0140, -0.0161], + [-0.0052, 0.0168, -0.0059, ..., -0.0092, 0.0124, 0.0043], + ..., + [ 0.0398, 0.0452, 0.0251, ..., 0.0039, -0.0330, 0.0051], + [-0.0620, -0.0039, 0.0371, ..., 0.0410, 0.0061, -0.0173], + [ 0.0127, 0.0074, -0.0270, ..., 0.0081, 0.0388, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0040, 0.0245], + [-0.0449, 0.0025, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0064, ..., 0.0046, 0.0288, 0.0125], + [ 0.0330, 0.0038, 0.0057, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0069, 0.0045, ..., -0.0077, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0070, 0.0038, 0.0527], + [-0.0051, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0095, -0.0096, 0.0327], + [ 0.0146, 0.0116, -0.0030, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0129, 0.0281, ..., -0.0242, -0.0126, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0155, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0041], + [-0.0048, -0.0045, -0.0286, ..., 0.0050, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0057, ..., -0.0019, 0.0082, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0096, 0.0150], + [-0.0071, 0.0029, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0063, ..., 0.0092, -0.0114, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0070, -0.0033, ..., -0.0220, -0.0145, -0.0046], + [-0.0166, 0.0189, -0.0058, ..., 0.0096, 0.0033, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0067, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0066, ..., 0.0024, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0069, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0121, 0.0289, 0.0126, ..., 0.0134, 0.0066, -0.0103], + [ 0.0188, -0.0197, -0.0022, ..., 0.0284, -0.0342, -0.0058], + [-0.0525, -0.0214, 0.0154, ..., -0.0270, 0.0081, 0.0254], + ..., + [-0.0052, 0.0087, 0.0281, ..., 0.0129, -0.0042, 0.0049], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0037, -0.0054], + [-0.0273, -0.0747, 0.0208, ..., -0.0042, -0.0052, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 9.1171e-04, 1.4954e-02, 3.3447e-02, ..., 2.0508e-02, + -2.7344e-02, -3.7537e-03], + [-1.6479e-02, 1.1108e-02, -1.3489e-02, ..., 1.0620e-02, + -7.2327e-03, 2.1851e-02], + [ 5.0659e-03, 2.2705e-02, -5.6458e-03, ..., 1.2817e-02, + -7.1106e-03, 1.4954e-02], + ..., + [-3.4668e-02, -1.4709e-02, 6.8359e-03, ..., 5.5847e-03, + -5.0659e-03, -9.2773e-03], + [-6.5613e-03, 3.9795e-02, -3.4424e-02, ..., 2.4536e-02, + 6.3477e-02, -1.1780e-02], + [-5.3024e-04, -5.0049e-02, -2.3193e-02, ..., 3.1982e-02, + -7.5340e-05, -5.5847e-03]], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0154, ..., -0.0344, 0.0067, -0.0364], + [ 0.0220, -0.0030, -0.0131, ..., 0.0078, 0.0101, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, 0.0072, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0073, 0.0192, -0.0393], + [-0.0105, -0.0244, 0.0275, ..., -0.0052, 0.0042, 0.0081], + [-0.0170, -0.0427, 0.0142, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0046, -0.0378, -0.0359, ..., 0.0552, -0.0078, -0.0160], + [ 0.0045, 0.0140, 0.0133, ..., 0.0050, -0.0153, 0.0354], + [ 0.0095, -0.0042, 0.0045, ..., -0.0047, -0.0249, -0.0136], + ..., + [ 0.0079, -0.0125, 0.0051, ..., -0.0095, -0.0084, 0.0052], + [-0.0060, 0.0439, 0.0166, ..., 0.0248, 0.0132, -0.0187], + [ 0.0175, 0.0063, 0.0044, ..., 0.0410, -0.0096, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0105, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0145, 0.0025, ..., 0.0065, -0.0244, 0.0356], + [ 0.0079, 0.0165, -0.0200, ..., -0.0080, 0.0237, -0.0081], + ..., + [-0.0197, -0.0022, 0.0386, ..., 0.0181, 0.0113, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, -0.0041, -0.0374], + [-0.0086, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0075, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0137, 0.0115, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, 0.0013, -0.0066, ..., 0.0050, -0.0045, 0.0337], + ..., + [ 0.0422, 0.0054, 0.0352, ..., -0.0304, -0.0035, -0.0204], + [ 0.0302, 0.0156, 0.0100, ..., -0.0049, -0.0254, -0.0245], + [-0.0067, -0.0064, 0.0078, ..., -0.0078, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0044, 0.0266, ..., 0.0154, 0.0187, 0.0049], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0096], + [-0.0189, 0.0205, 0.0044, ..., -0.0172, -0.0090, 0.0231], + ..., + [-0.0295, -0.0137, 0.0045, ..., 0.0039, -0.0050, 0.0099], + [-0.0052, -0.0140, 0.0317, ..., -0.0153, -0.0459, -0.0065], + [-0.0041, -0.0051, -0.0259, ..., -0.0040, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0059, -0.0140, ..., 0.0242, 0.0055, -0.0386], + [-0.0141, 0.0060, -0.0097, ..., -0.0369, -0.0155, 0.0157], + [-0.0267, -0.0041, -0.0065, ..., -0.0127, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0029, 0.0144, ..., 0.0115, 0.0040, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., 0.0045, -0.0038, -0.0137], + [-0.0219, -0.0396, -0.0010, ..., -0.0077, -0.0033, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0051, 0.0168, 0.0194, ..., 0.0349, -0.0420, -0.0048], + [ 0.0044, 0.0273, 0.0146, ..., -0.0184, 0.0011, 0.0156], + [ 0.0145, -0.0031, -0.0070, ..., 0.0245, 0.0359, 0.0061], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0037, 0.0043, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0072, -0.0312, -0.0457], + [ 0.0055, 0.0039, -0.0058, ..., -0.0038, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0047, 0.0121, 0.0038, ..., 0.0124, 0.0006, 0.0067], + [-0.0277, -0.0039, -0.0181, ..., 0.0005, 0.0020, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0038, 0.0513, -0.0113], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0113], + [-0.0282, -0.0040, 0.0121, ..., -0.0057, -0.0039, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, 0.0048, -0.0128, ..., 0.0140, 0.0308, -0.0066], + [ 0.0120, -0.0332, -0.0320, ..., -0.0045, -0.0140, -0.0039], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0088, 0.0295, 0.0505], + [-0.0159, 0.0025, -0.0311, ..., 0.0120, 0.0063, 0.0081], + [ 0.0044, 0.0198, 0.0141, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, 0.0043, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0090, -0.0040, 0.0052, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0045, ..., 0.0302, -0.0037, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0041, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0020, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0045, -0.0222, 0.0262], + ..., + [-0.0147, 0.0051, 0.0260, ..., 0.0045, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0057, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0042, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0112], + [-0.0278, -0.0054, -0.0374, ..., 0.0276, -0.0046, 0.0151], + [ 0.0089, 0.0237, -0.0076, ..., 0.0527, -0.0078, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0075, -0.0095], + [-0.0068, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0039, -0.0128, ..., 0.0060, -0.0371, -0.0046], + [ 0.0071, -0.0127, 0.0045, ..., -0.0204, 0.0039, 0.0049], + [-0.0110, -0.0195, -0.0050, ..., -0.0542, -0.0092, -0.0063], + ..., + [-0.0205, -0.0084, 0.0039, ..., -0.0199, 0.0046, 0.0039], + [-0.0220, -0.0302, -0.0101, ..., 0.0105, 0.0369, 0.0041], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[-0.0042, 0.0347, 0.0092, ..., 0.0277, -0.0208, -0.0220], + [-0.0038, 0.0043, 0.0238, ..., -0.0145, -0.0320, -0.0046], + [-0.0193, 0.0364, 0.0039, ..., -0.0077, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0065, -0.0232, -0.0464], + [-0.0110, -0.0298, -0.0195, ..., 0.0112, 0.0312, 0.0154], + [-0.0072, -0.0119, -0.0195, ..., 0.0041, 0.0109, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0078, -0.0239, -0.0123, ..., -0.0062, 0.0243, -0.0396], + [-0.0100, 0.0154, 0.0026, ..., -0.0277, -0.0027, -0.0038], + ..., + [ 0.0117, -0.0052, 0.0515, ..., 0.0327, -0.0120, -0.0074], + [ 0.0039, 0.0170, 0.0055, ..., 0.0039, 0.0280, -0.0210], + [-0.0025, 0.0072, 0.0045, ..., 0.0155, 0.0042, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [ 0.0040, 0.0081, -0.0074, ..., -0.0588, 0.0187, -0.0078], + [-0.0102, 0.0199, -0.0078, ..., 0.0080, -0.0332, -0.0076], + ..., + [-0.0052, -0.0464, 0.0277, ..., -0.0143, 0.0198, 0.0054], + [ 0.0162, 0.0011, 0.0085, ..., -0.0104, -0.0066, 0.0297], + [-0.0269, 0.0131, 0.0164, ..., 0.0084, 0.0176, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0046, ..., -0.0505, -0.0332, 0.0154], + [ 0.0208, 0.0366, 0.0007, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, 0.0058, -0.0201, ..., 0.0330, 0.0309, -0.0064], + ..., + [ 0.0028, 0.0123, 0.0153, ..., 0.0039, 0.0034, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0042, 0.0175], + [-0.0244, -0.0062, -0.0300, ..., -0.0061, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, 0.0040, 0.0075, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0050, ..., 0.0310, 0.0088, -0.0039], + [-0.0078, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0087, -0.0109, -0.0237, ..., -0.0104, 0.0157, -0.0021], + [-0.0085, 0.0205, 0.0145, ..., 0.0262, -0.0195, 0.0095], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0070]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0029, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0029, 0.0034, -0.0093], + [ 0.0017, -0.0011, 0.0398, ..., -0.0304, -0.0030, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0039], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0048, 0.0291], + [ 0.0132, -0.0004, -0.0339, ..., 0.0065, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, 0.0024], + [-0.0166, -0.0043, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0064, 0.0020, ..., 0.0240, 0.0031, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0020], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0009, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0066, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0041], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, 0.0034, 0.0021], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, -0.0042, ..., -0.0164, -0.0040, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0027, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0053, -0.0231, -0.0255, ..., -0.0068, 0.0295, -0.0046], + [-0.0255, 0.0513, 0.0030, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0035, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0064, -0.0066, -0.0364, ..., -0.0175, 0.0306, -0.0077], + [-0.0051, -0.0195, -0.0027, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0078, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0078, -0.0193, -0.0140], + [-0.0166, -0.0078, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0134, -0.0044, 0.0137, ..., 0.0021, 0.0211, -0.0275], + [ 0.0051, -0.0076, -0.0234, ..., 0.0088, -0.0084, -0.0166], + [ 0.0225, -0.0251, 0.0039, ..., 0.0072, 0.0236, 0.0077], + ..., + [ 0.0184, 0.0145, -0.0041, ..., 0.0269, -0.0085, -0.0096], + [-0.0038, 0.0044, 0.0078, ..., -0.0260, 0.0267, 0.0150], + [ 0.0102, 0.0043, 0.0212, ..., 0.0195, -0.0157, 0.0124]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0085, -0.0270, 0.0251, ..., 0.0095, 0.0052, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.0.proj.1.weight': tensor([[ 0.0018, 0.0251, 0.0957, ..., -0.0535, 0.0286, -0.0262], + [-0.0195, 0.0112, -0.1108, ..., 0.0304, -0.0312, 0.0593], + [ 0.0017, 0.1309, -0.0078, ..., -0.0058, 0.0078, -0.0282], + ..., + [ 0.0039, -0.0038, -0.0136, ..., 0.0142, -0.0222, -0.0051], + [-0.0349, -0.0645, 0.0508, ..., 0.0364, 0.0630, 0.0566], + [ 0.0192, -0.0223, 0.0010, ..., 0.0115, 0.0510, -0.0084]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_proj.weight': tensor([[ 0.0182, 0.0081, 0.0123, ..., 0.0016, -0.0015, 0.0272], + [ 0.0183, -0.0024, 0.0045, ..., 0.0009, -0.0028, 0.0209], + [-0.0061, 0.0053, -0.0056, ..., -0.0055, -0.0039, -0.0149], + ..., + [-0.0089, -0.0040, -0.0085, ..., 0.0106, -0.0037, -0.0076], + [-0.0016, -0.0033, -0.0052, ..., -0.0043, -0.0039, -0.0052], + [ 0.0081, 0.0026, 0.0023, ..., 0.0082, -0.0038, 0.0005]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_norm.weight': tensor([ 1.0742e-01, -9.8633e-02, 1.1279e-01, 3.6914e-01, -4.8056e-07, + -1.9092e-07, 2.0752e-03, 7.8870e-13, 5.4956e-15, 4.9738e-13, + -2.0961e-13, 3.3307e-16, 1.1519e-15, -4.4631e-14, 1.0868e-10, + -4.3485e-12, -3.3617e-05, 1.4827e-06, -8.6212e-04, 5.6744e-05, + 1.1505e-10, -2.5466e-11, 1.9409e-02, -4.1962e-04, 3.0518e-02, + 6.6406e-01, 1.3867e-01, -2.4414e-03, 5.6641e-02, 1.0596e-01, + 1.3279e-10, 5.4845e-14, -2.2736e-03, 1.4663e-05, 2.9755e-03, + 1.0376e-03, 6.8212e-11, -2.9650e-10, 4.8637e-05, 1.5099e-14, + 9.0639e-17, -1.2434e-14, 6.4393e-15, 4.4250e-03, 3.2471e-02, + 2.2125e-03, -3.5400e-02, -4.6875e-02, 4.8242e-01, 2.6703e-03, + -6.8665e-03, 3.4027e-03, 3.9062e-03, 1.6838e-06, 2.1210e-03, + 1.2733e-10, -1.3131e-11, 3.3396e-13, -4.5103e-16, 1.0249e-19, + -1.2967e-13, -3.8858e-15, 2.6557e-10, 3.3168e-15, 3.3203e-01, + 3.4766e-01, 4.3750e-01, 1.1768e-01, -2.4959e-07, -1.0245e-07, + 6.9580e-03, 1.5135e-12, -2.9088e-14, -3.2152e-13, 2.8866e-14, + 3.3307e-16, -1.2663e-16, -2.4425e-14, 1.8758e-11, -6.3949e-13, + 7.3910e-05, 4.0770e-05, 2.8133e-05, 5.7220e-05, -5.3478e-10, + -1.2296e-09, -8.8120e-04, -1.3580e-03, 4.8633e-01, 9.0820e-02, + 6.0938e-01, 7.9590e-02, 4.5471e-03, -2.6550e-03, 5.6389e-11, + 1.0081e-13, 2.9373e-04, 1.6117e-04, -9.7752e-06, 4.9210e-04, + -5.7116e-10, 1.4734e-10, 3.7994e-03, -3.6915e-15, 7.5287e-16, + -4.1980e-16, 2.0400e-15, 2.5269e-02, -6.6223e-03, -1.1978e-03, + 4.5117e-01, 2.8516e-01, 1.0498e-01, -7.8125e-03, 7.8125e-03, + -2.9907e-03, -4.0894e-03, 2.6375e-06, 2.1057e-03, -1.4552e-10, + 3.4333e-11, -4.5119e-13, 1.4502e-15, -3.6321e-18, 4.6896e-13, + -9.6367e-14, 4.2917e-12, 3.7303e-14], dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.self_attn.k_proj.weight': tensor([[-0.0102, 0.0103, -0.0096, ..., -0.0046, -0.0105, 0.0022], + [-0.0124, 0.0089, 0.0037, ..., 0.0029, -0.0098, 0.0126], + [-0.0073, -0.0295, 0.0043, ..., -0.0023, 0.0043, -0.0208], + ..., + [-0.0018, 0.0002, -0.0042, ..., -0.0025, -0.0005, -0.0027], + [-0.0027, -0.0030, -0.0047, ..., 0.0026, 0.0041, 0.0038], + [-0.0037, -0.0039, -0.0026, ..., 0.0029, -0.0006, 0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.k_norm.weight': tensor([ 2.9102e-01, 2.0312e-01, -2.0508e-01, 3.4961e-01, 4.7497e-07, + -2.1793e-07, 4.3335e-03, 8.1357e-13, 1.1990e-14, 5.0449e-13, + 1.8741e-13, 2.9837e-16, 1.1519e-15, -4.4853e-14, -1.0823e-10, + -4.4054e-12, 7.4863e-05, -1.2945e-07, 1.0910e-03, 5.6744e-05, + 1.2187e-10, -2.9786e-11, 1.3367e-02, 7.2861e-04, 4.1260e-02, + 6.6406e-01, 1.9336e-01, 7.6660e-02, 4.1809e-03, 2.9206e-05, + -1.3733e-10, 6.8390e-14, 2.2888e-03, 4.6492e-05, 2.9907e-03, + 1.0376e-03, 7.9126e-11, 2.9286e-10, 6.8665e-05, 1.4322e-14, + 1.1241e-15, -6.6058e-15, -6.4670e-15, 2.1118e-02, 2.9297e-02, + 2.1820e-03, 1.0254e-01, 1.1963e-01, 5.0391e-01, 7.8125e-03, + -7.8125e-03, -3.9062e-03, 4.0894e-03, -3.2783e-06, -2.0142e-03, + -6.5938e-11, -1.4353e-12, 3.4639e-13, 2.3419e-16, -4.7095e-19, + 1.4655e-13, -7.8271e-15, -2.6739e-10, 3.7470e-15, 1.7285e-01, + 3.4180e-01, 4.3164e-01, -2.6172e-01, -2.3562e-07, -1.1583e-08, + 3.5553e-03, 1.4566e-12, -2.6978e-14, 3.3751e-13, 3.1752e-14, + -3.8858e-16, 1.0278e-16, 1.5543e-14, -1.9554e-11, -6.6436e-13, + 1.1349e-04, -4.7684e-05, 1.4603e-05, 5.8413e-05, 5.3842e-10, + -1.2151e-09, 2.1744e-04, -9.0408e-04, 4.8242e-01, -8.9355e-02, + 6.0156e-01, -3.7689e-03, 5.5664e-02, 1.0547e-01, 1.1880e-11, + 2.2093e-14, -3.5095e-04, 1.6117e-04, -1.9169e-04, -4.8256e-04, + 5.7844e-10, -2.5466e-10, 3.7689e-03, 3.7748e-15, 5.4678e-15, + -8.6736e-17, 2.0817e-15, 1.4282e-02, 1.7334e-02, 1.4725e-03, + 4.5312e-01, 2.6172e-01, 1.4453e-01, 4.6082e-03, 7.6294e-03, + 1.3199e-03, 3.0975e-03, 9.3877e-07, 2.2583e-03, 1.9281e-10, + 3.6607e-11, 4.1389e-13, -1.4919e-15, 3.5508e-18, -4.6896e-13, + -9.6811e-14, 1.1084e-12, 3.7303e-14], dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.self_attn.v_proj.weight': tensor([[-0.0022, 0.0098, 0.0055, ..., -0.0032, -0.0040, -0.0063], + [ 0.0002, 0.0038, -0.0006, ..., -0.0014, 0.0046, 0.0002], + [ 0.0142, 0.0043, 0.0038, ..., -0.0041, 0.0016, 0.0063], + ..., + [-0.0006, 0.0021, -0.0019, ..., 0.0025, -0.0025, -0.0029], + [ 0.0063, -0.0032, 0.0037, ..., 0.0059, -0.0058, 0.0040], + [-0.0090, 0.0049, -0.0024, ..., 0.0055, -0.0038, -0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.output_proj.weight': tensor([[ 0.0145, -0.0216, -0.0175, ..., 0.0042, 0.0087, -0.0032], + [ 0.0022, -0.0068, 0.0124, ..., -0.0063, -0.0024, -0.0059], + [ 0.0040, 0.0022, 0.0052, ..., 0.0009, -0.0058, 0.0017], + ..., + [ 0.0055, -0.0001, 0.0048, ..., 0.0042, -0.0014, 0.0009], + [-0.0055, -0.0045, 0.0275, ..., 0.0036, -0.0033, 0.0161], + [-0.0084, 0.0095, -0.0052, ..., 0.0051, 0.0051, -0.0053]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_proj.weight': tensor([[-0.0034, -0.0093, -0.0017, ..., -0.0048, -0.0034, -0.0014], + [-0.0096, 0.0098, 0.0019, ..., 0.0072, 0.0025, -0.0030], + [-0.0023, -0.0084, 0.0025, ..., -0.0004, 0.0062, 0.0029], + ..., + [ 0.0030, 0.0071, 0.0009, ..., 0.0047, -0.0041, 0.0082], + [-0.0015, -0.0042, -0.0017, ..., -0.0048, 0.0043, -0.0067], + [-0.0060, 0.0023, 0.0009, ..., -0.0008, 0.0033, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_norm.weight': tensor([ 6.0120e-03, 8.8811e-06, 2.7740e-11, 1.4526e-02, -9.6436e-03, + 1.5442e-02, 2.2054e-06, 5.4550e-04, 3.4912e-02, 2.6978e-02, + 1.6880e-08, 2.8076e-02, 5.7812e-01, 9.3994e-03, 1.0437e-02, + 4.0894e-03, 4.6680e-01, 1.1086e-05, 1.1215e-03, 1.0803e-02, + 8.8501e-03, 7.3547e-03, 8.3496e-02, 1.8120e-05, 1.8433e-02, + 1.0681e-03, 7.9346e-03, -8.4400e-05, 1.2436e-03, 1.6708e-03, + 3.5156e-01, 9.0942e-03, 5.0306e-05, 4.6680e-01, 5.4297e-01, + 1.3828e-04, -5.1022e-05, 3.1494e-02, 1.5137e-01, 2.5635e-02, + 2.0996e-02, -4.6790e-06, 1.3642e-10, 4.1809e-03, 9.4223e-04, + 3.3203e-02, 1.2024e-02, 1.0605e-03, 4.0894e-03, 8.7891e-03, + 1.8539e-03, -7.4863e-05, -4.2915e-05, 1.5106e-03, -3.8300e-03, + 1.6689e-04, 1.9653e-02, 1.1396e-04, 1.7090e-02, 1.1475e-02, + 1.9932e-04, -7.8082e-06, 1.2329e-02, 7.2632e-03, 4.0293e-05, + 1.3550e-02, 9.9945e-04, 2.5000e-01, 8.8501e-04, 2.1851e-02, + 4.6875e-01, -1.2875e-05, 4.5410e-02, 8.4229e-03, 1.7695e-08, + 7.0190e-03, 1.6632e-03, 7.0801e-03, -1.3590e-05, 2.0752e-02, + 2.3438e-02, 4.2725e-03, 1.7188e-01, 7.6294e-03, 1.5869e-03, + 2.4567e-03, 1.3000e-02, -1.3530e-05, 1.0395e-04, 5.0781e-01, + 2.6550e-03, 8.9645e-05, 1.1520e-03, 3.8147e-06, 9.3384e-03, + 4.5117e-01, 4.3640e-03, 1.0538e-04, -3.7126e-13, 9.4223e-04, + 3.9062e-01, 1.3351e-05, -9.0332e-03, 1.7822e-02, 2.3071e-02, + 9.3460e-05, -1.5080e-05, 1.3769e-05, 1.6708e-03, 1.7944e-02, + 1.1597e-02, 8.5449e-04, 4.2419e-03, 1.2329e-02, -3.6316e-03, + 1.0395e-04, 1.1230e-02, -2.8534e-03, -1.1292e-03, 1.8311e-03, + 1.0300e-04, 6.3782e-03, 1.1368e-03, -2.2650e-05, 1.8978e-04, + 1.1414e-02, -7.1526e-06, 1.0133e-05], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_proj.weight': tensor([[ 1.5411e-03, 7.2861e-04, 1.9150e-03, ..., -6.7520e-04, + -2.4605e-04, 6.1035e-04], + [-3.6926e-03, -2.0599e-03, -3.9368e-03, ..., -2.6093e-03, + -3.9673e-03, -3.0060e-03], + [-1.7700e-03, -1.4877e-03, -2.4872e-03, ..., -2.8381e-03, + -5.6458e-03, -9.6130e-04], + ..., + [ 2.5635e-03, -2.4719e-03, 1.8768e-03, ..., 5.2643e-04, + 2.3041e-03, -5.7220e-04], + [-8.8882e-04, 5.3024e-04, -2.3193e-03, ..., -3.0212e-03, + 5.0354e-04, 2.3193e-03], + [ 2.9297e-03, 1.8997e-03, -3.2501e-03, ..., -4.2725e-03, + 6.5613e-04, 5.7697e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_norm.weight': tensor([ 5.9814e-03, -5.5730e-06, -2.7740e-11, 1.4343e-02, 9.7046e-03, + 1.5442e-02, 2.1756e-06, -5.4550e-04, 3.3936e-02, 2.7100e-02, + -5.9954e-09, 2.8320e-02, 5.8203e-01, 9.3384e-03, 9.8877e-03, + 4.0588e-03, 4.6680e-01, 1.1325e-05, 9.9182e-04, 1.0620e-02, + 8.7280e-03, 7.3853e-03, 8.3496e-02, 3.7670e-05, 1.8311e-02, + 1.0757e-03, 7.5073e-03, -8.3923e-05, 6.2561e-04, 1.6403e-03, + 3.5156e-01, 8.9111e-03, 5.0068e-05, 4.6680e-01, 5.4297e-01, + -2.0981e-04, 5.8413e-05, 3.1982e-02, 1.5137e-01, 2.5879e-02, + 2.0996e-02, 9.7275e-05, 1.9936e-09, 4.0283e-03, -2.2793e-04, + 3.2715e-02, 1.1902e-02, 1.0757e-03, 4.0283e-03, 8.7280e-03, + 1.8463e-03, -7.4863e-05, 4.2915e-05, 1.5030e-03, 3.9978e-03, + 2.1076e-04, 1.9531e-02, 1.3351e-04, 1.7334e-02, 1.1536e-02, + 1.9932e-04, -8.1062e-06, 1.2085e-02, 7.2632e-03, 4.0054e-05, + 1.3733e-02, 9.9182e-04, 2.5000e-01, 8.9645e-04, 2.1729e-02, + 4.6875e-01, 1.2696e-05, 4.3945e-02, 8.4229e-03, -1.9674e-08, + 7.1411e-03, 5.5313e-04, 7.0190e-03, -1.4782e-05, 2.0752e-02, + 2.3560e-02, 4.2419e-03, 1.7188e-01, 7.4463e-03, 1.6251e-03, + 2.4872e-03, 1.4038e-02, 1.3590e-05, 1.0490e-04, 5.0781e-01, + 2.6398e-03, -1.2398e-04, 7.5912e-04, 3.7253e-06, 9.3384e-03, + 4.5117e-01, -4.2419e-03, -9.1553e-05, 3.7126e-13, 9.1934e-04, + 3.9062e-01, 1.3351e-05, 1.0193e-02, 1.7822e-02, 2.3071e-02, + 9.2983e-05, 1.7524e-05, 1.3888e-05, 8.1253e-04, 1.7944e-02, + 1.1841e-02, 8.5449e-04, 4.2114e-03, 1.2268e-02, 3.6163e-03, + -1.6785e-04, 1.1292e-02, 2.8687e-03, 1.3123e-03, 1.8845e-03, + 9.7275e-05, 6.5918e-03, 1.1139e-03, -9.0599e-05, 1.0633e-04, + 1.1292e-02, -6.8247e-06, -3.7700e-06], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.v_proj.weight': tensor([[-9.8877e-03, 2.1820e-03, 1.2695e-02, ..., 2.8534e-03, + 2.6550e-03, -1.0193e-02], + [-2.2461e-02, 8.1635e-04, 7.7209e-03, ..., -3.2043e-03, + -6.1035e-03, 9.3937e-05], + [-1.2573e-02, 2.4567e-03, 2.9297e-03, ..., -2.2888e-03, + 1.3000e-02, -1.2268e-02], + ..., + [ 1.0925e-02, 1.8692e-03, 1.6861e-03, ..., 2.5940e-04, + 2.9755e-03, 1.8799e-02], + [-9.5215e-03, 2.5330e-03, 1.0315e-02, ..., 1.8463e-03, + -1.7700e-02, -3.7079e-03], + [-1.6479e-02, -2.8839e-03, 7.1716e-03, ..., -2.0599e-03, + -6.0425e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.output_proj.weight': tensor([[ 0.0064, -0.0018, 0.0056, ..., -0.0030, 0.0018, 0.0080], + [ 0.0080, 0.0066, -0.0012, ..., 0.0162, 0.0041, -0.0110], + [ 0.0052, -0.0030, -0.0008, ..., -0.0004, -0.0073, -0.0052], + ..., + [-0.0071, -0.0077, -0.0051, ..., -0.0079, -0.0047, -0.0085], + [ 0.0101, -0.0025, 0.0059, ..., -0.0142, -0.0183, 0.0055], + [ 0.0106, -0.0081, -0.0041, ..., -0.0070, -0.0037, -0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_img.weight': tensor([[ 0.0067, 0.0119, -0.0041, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0075, 0.0249, 0.0291], + [-0.0095, 0.0046, -0.0115, ..., -0.0342, -0.0327, 0.0061], + ..., + [ 0.0074, -0.0067, 0.0040, ..., 0.0082, 0.0286, -0.0165], + [ 0.0078, 0.0013, -0.0102, ..., -0.0100, 0.0142, -0.0309], + [ 0.0242, -0.0106, 0.0255, ..., -0.0227, -0.0043, -0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, -0.0010], + [ 0.0026, 0.0120, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0085, ..., -0.0057, -0.0046, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0039, 0.0054], + [-0.0160, 0.0055, -0.0036, ..., 0.0133, -0.0173, -0.0049], + [-0.0051, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img.weight': tensor([[-0.0510, -0.0021, -0.0364, ..., -0.0225, -0.0143, -0.0055], + [-0.0157, 0.0053, -0.0364, ..., 0.0099, 0.0117, -0.0036], + [ 0.0025, 0.0615, 0.0056, ..., -0.0205, -0.0306, -0.0114], + ..., + [ 0.0008, -0.0018, -0.0103, ..., -0.0211, -0.0012, 0.0288], + [ 0.0266, -0.0115, -0.0056, ..., -0.0420, -0.0282, -0.0008], + [ 0.0286, 0.0325, 0.0095, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.mlp.layer1.weight': tensor([[ 0.0082, 0.0159, 0.0067, ..., 0.0008, 0.0096, 0.0064], + [ 0.0050, -0.0152, -0.0035, ..., -0.0021, -0.0115, 0.0064], + [ 0.0062, -0.0057, -0.0039, ..., 0.0176, -0.0176, 0.0121], + ..., + [-0.0037, 0.0004, 0.0240, ..., 0.0054, 0.0075, 0.0029], + [-0.0030, 0.0186, 0.0068, ..., 0.0021, -0.0079, 0.0085], + [ 0.0050, 0.0027, -0.0013, ..., 0.0029, -0.0074, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.mlp.layer2.weight': tensor([[-0.0077, -0.0042, -0.0074, ..., -0.0038, -0.0261, -0.0078], + [ 0.0042, 0.0085, -0.0029, ..., -0.0162, -0.0216, 0.0073], + [-0.0062, 0.0092, 0.0090, ..., 0.0012, -0.0023, 0.0053], + ..., + [-0.0044, -0.0049, -0.0091, ..., -0.0069, -0.0026, -0.0082], + [ 0.0079, 0.0125, 0.0215, ..., 0.0008, 0.0056, 0.0114], + [ 0.0085, 0.0122, 0.0012, ..., 0.0050, -0.0021, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0610, -0.0593, -0.0579, ..., 0.0030, 0.0031, 0.0031], + [ 0.0591, 0.0583, 0.0562, ..., -0.0024, -0.0025, -0.0026], + [ 0.0583, 0.0576, 0.0564, ..., -0.0026, -0.0027, -0.0027], + ..., + [ 0.0630, 0.0615, 0.0598, ..., -0.0024, -0.0025, -0.0025], + [ 0.0598, 0.0583, 0.0566, ..., -0.0022, -0.0023, -0.0022], + [ 0.0598, 0.0583, 0.0571, ..., -0.0030, -0.0031, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0009, 0.0014, 0.0015, ..., 0.0050, 0.0036, 0.0003], + [-0.0025, -0.0015, -0.0016, ..., 0.0022, 0.0022, 0.0024], + [-0.0035, 0.0033, 0.0031, ..., 0.0059, 0.0060, 0.0034], + ..., + [-0.0432, 0.0500, 0.0493, ..., 0.0505, 0.0464, 0.0437], + [-0.0058, 0.0069, 0.0061, ..., 0.0053, 0.0044, 0.0057], + [ 0.0096, -0.0104, -0.0103, ..., -0.0099, -0.0096, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_cross_attn.1.weight': tensor([[-8.5449e-03, 4.4556e-03, 4.4861e-03, ..., -6.7520e-04, + -6.0654e-04, -7.7438e-04], + [-3.8818e-02, -3.7109e-02, -3.7109e-02, ..., 2.5177e-04, + 2.5940e-04, 2.6512e-04], + [ 4.1748e-02, 4.0283e-02, 4.0039e-02, ..., -7.0095e-05, + -5.1022e-05, -7.5340e-05], + ..., + [ 3.9062e-02, 3.8086e-02, 3.7109e-02, ..., -3.3569e-04, + -2.6703e-04, -3.0327e-04], + [ 4.9072e-02, 4.8096e-02, 4.6631e-02, ..., 5.0354e-04, + 6.1417e-04, 5.4550e-04], + [ 3.8818e-02, 3.7598e-02, 3.6621e-02, ..., -5.6624e-06, + 7.0572e-05, 2.1935e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0129, -0.0072, 0.0067, ..., 0.0072, 0.0072, 0.0068], + [ 0.0052, 0.0034, -0.0049, ..., -0.0034, -0.0034, -0.0048], + [-0.0037, -0.0036, 0.0035, ..., 0.0036, 0.0035, 0.0036], + ..., + [-0.0063, -0.0205, 0.0364, ..., 0.0203, 0.0432, 0.0320], + [ 0.0103, 0.0139, 0.0101, ..., -0.0131, 0.0050, 0.0099], + [-0.0137, -0.0046, -0.0113, ..., 0.0045, -0.0094, -0.0110]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_mlp.1.weight': tensor([[-0.0049, -0.0049, 0.0038, ..., 0.0004, 0.0007, 0.0006], + [ 0.0640, 0.0618, 0.0605, ..., -0.0024, -0.0027, -0.0027], + [ 0.0554, 0.0542, 0.0525, ..., -0.0037, -0.0040, -0.0040], + ..., + [ 0.0952, 0.0928, 0.0908, ..., -0.0008, -0.0009, -0.0010], + [-0.0038, -0.0038, -0.0038, ..., -0.0004, -0.0005, -0.0004], + [ 0.0747, 0.0723, 0.0708, ..., -0.0012, -0.0012, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_mlp.2.weight': tensor([[-0.0077, 0.0091, 0.0107, ..., 0.0072, 0.0082, 0.0060], + [-0.0007, 0.0023, 0.0033, ..., 0.0031, 0.0007, 0.0009], + [-0.0017, 0.0060, 0.0061, ..., 0.0044, 0.0015, 0.0057], + ..., + [ 0.0036, 0.0747, 0.1025, ..., 0.0635, -0.0044, 0.0496], + [ 0.0114, 0.0165, 0.0096, ..., 0.0280, -0.0114, 0.0233], + [-0.0089, 0.0033, -0.0003, ..., 0.0041, 0.0087, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.before_proj.weight': tensor([[ 0.0028, 0.0002, -0.0006, ..., 0.0062, 0.0011, -0.0013], + [-0.0094, -0.0055, -0.0041, ..., -0.0078, -0.0024, 0.0052], + [-0.0075, -0.0059, -0.0033, ..., -0.0015, -0.0036, -0.0027], + ..., + [-0.0011, 0.0017, -0.0047, ..., 0.0033, -0.0049, -0.0004], + [-0.0037, -0.0017, 0.0011, ..., -0.0020, -0.0052, -0.0042], + [ 0.0027, 0.0028, -0.0021, ..., 0.0045, 0.0005, -0.0034]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.before_proj.bias': tensor([-0.0199, 0.0027, 0.0018, ..., 0.0016, -0.0040, -0.0239], + dtype=torch.bfloat16), 'net.control_blocks_0.0.after_proj.weight': tensor([[-0.0188, -0.0083, 0.0017, ..., -0.0049, -0.0119, 0.0113], + [ 0.0189, -0.0134, 0.0063, ..., -0.0007, 0.0084, -0.0087], + [ 0.0009, -0.0009, -0.0043, ..., -0.0043, 0.0015, -0.0023], + ..., + [ 0.0029, -0.0039, 0.0083, ..., -0.0016, -0.0035, 0.0015], + [ 0.0276, -0.0150, 0.0077, ..., -0.0062, -0.0171, -0.0134], + [-0.0036, -0.0123, 0.0019, ..., 0.0026, -0.0048, 0.0112]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.after_proj.bias': tensor([ 0.0075, -0.0165, -0.0026, ..., -0.0014, -0.0140, 0.0598], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_proj.weight': tensor([[-0.0220, 0.0138, -0.0125, ..., -0.0093, -0.0069, 0.0132], + [ 0.0049, 0.0032, -0.0035, ..., 0.0193, -0.0050, -0.0388], + [ 0.0012, 0.0103, 0.0013, ..., -0.0023, 0.0043, -0.0078], + ..., + [ 0.0009, 0.0050, -0.0006, ..., 0.0024, -0.0010, -0.0002], + [ 0.0021, -0.0055, -0.0013, ..., 0.0013, -0.0062, 0.0017], + [ 0.0025, -0.0006, 0.0052, ..., 0.0031, -0.0019, -0.0037]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_norm.weight': tensor([ 3.9062e-01, 1.2578e+00, 5.6396e-02, 1.1953e+00, 8.9844e-01, + 4.6945e-34, -1.9184e-34, -1.5573e-34, -6.8310e-34, 3.3087e-22, + 8.9676e-34, -4.9653e-34, -9.3287e-35, -3.7014e-34, -7.2523e-34, + -1.0653e-33, 1.4551e-01, 3.8338e-04, -2.1696e-05, 2.2339e-02, + 2.4319e-04, -5.7697e-05, 4.6289e-01, 8.8672e-01, 1.3516e+00, + 4.4531e-01, -8.6426e-02, 6.0547e-01, 8.4229e-03, -2.5022e-30, + 1.9259e-34, 1.6852e-33, -2.3127e-05, 8.9264e-04, 8.7738e-05, + -9.7656e-03, 3.1281e-04, -1.0909e-34, -4.9277e-35, 1.1977e-33, + -1.3542e-33, 2.1507e-22, 6.3496e-34, 1.0859e+00, 3.2422e-01, + 2.1973e-01, 3.7500e-01, 1.3438e+00, 2.0630e-02, -4.4556e-03, + -6.2256e-03, 2.3926e-01, 1.7395e-03, -4.6042e-34, -3.7917e-34, + -3.0333e-33, 7.4158e-03, -6.0100e-25, 1.9019e-33, 2.0613e-34, + 3.8556e-35, 4.0813e-35, -1.6325e-34, -2.7836e-35, 5.0391e-01, + 4.1406e-01, 6.3281e-01, 4.9805e-01, 1.6797e-01, 5.2662e-34, + -1.2883e-35, 4.4838e-34, 7.4028e-34, -3.2467e-23, 2.1667e-34, + -1.5528e-33, 2.7836e-35, -1.0833e-34, 9.3287e-34, 8.1852e-34, + 4.6387e-03, 8.9264e-04, -4.4346e-05, 4.4250e-04, 2.3079e-04, + 3.5889e-02, 1.1953e+00, 2.5391e-01, 4.5117e-01, 1.3359e+00, + 2.7734e-01, 2.4414e-01, 1.9455e-03, -2.0338e-30, 2.1215e-34, + -1.3120e-33, -3.0279e-05, 3.4714e-04, -1.6212e-05, 1.3550e-02, + 3.8719e-04, -1.1887e-34, -4.1001e-35, -1.6777e-34, -1.1857e-33, + -4.1359e-24, 4.2130e-33, 2.8711e-01, 9.4141e-01, 1.2266e+00, + 1.4844e+00, 1.9336e-01, 8.3594e-01, -3.9062e-03, -8.2397e-04, + -4.7913e-03, 2.5558e-04, 1.3181e-33, -3.4005e-34, 2.1787e-33, + 6.7520e-04, -7.9616e-24, 2.3232e-33, 1.7379e-34, -1.0909e-35, + 1.1134e-34, -6.9213e-35, 6.3947e-35], dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.self_attn.k_proj.weight': tensor([[ 0.0016, -0.0195, 0.0078, ..., 0.0172, -0.0051, -0.0264], + [ 0.0066, -0.0129, -0.0035, ..., -0.0199, 0.0034, -0.0095], + [ 0.0092, 0.0063, -0.0018, ..., 0.0054, -0.0044, 0.0065], + ..., + [-0.0085, 0.0008, -0.0060, ..., 0.0082, 0.0109, -0.0054], + [-0.0095, -0.0052, 0.0011, ..., -0.0036, 0.0062, -0.0051], + [-0.0058, -0.0028, -0.0025, ..., -0.0032, 0.0117, -0.0054]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.k_norm.weight': tensor([ 5.0391e-01, 1.2812e+00, 2.3438e-01, 1.2109e+00, 8.9844e-01, + 4.5139e-34, 1.3918e-34, 1.1586e-34, 1.5167e-33, 3.3087e-22, + -2.8137e-34, -2.0162e-34, 4.4387e-35, 1.9410e-34, -6.7708e-34, + 6.0938e-35, 1.4551e-01, 3.6812e-04, 1.2875e-04, 5.5420e-02, + 2.4319e-04, 1.2493e-04, 6.8750e-01, 8.9062e-01, 1.3516e+00, + 4.5117e-01, 2.1289e-01, 6.4453e-01, -8.8501e-03, 1.6147e-30, + 1.0382e-34, -7.8993e-35, 1.4782e-04, 8.6975e-04, 1.5545e-04, + 4.5410e-02, 3.0708e-04, -5.0781e-35, -2.4262e-35, 4.1829e-34, + -3.9120e-34, -8.5200e-23, 4.7245e-34, 1.0781e+00, 3.8672e-01, + -6.3477e-02, -3.8672e-01, 1.3359e+00, 1.5527e-01, 2.4567e-03, + -6.1646e-03, 2.3340e-01, 1.7395e-03, -2.5278e-34, 1.2940e-34, + 1.1857e-33, 7.1411e-03, -6.0746e-25, -3.3102e-34, 1.3015e-34, + -1.2413e-35, -6.1314e-35, 1.6852e-34, -2.2569e-35, 5.1562e-01, + 5.5078e-01, 6.2891e-01, 4.3359e-01, -2.4414e-01, -6.9514e-34, + 1.1360e-34, 2.0162e-34, -3.6352e-33, 1.5406e-23, -2.9190e-34, + -4.9051e-34, 1.9184e-35, 8.0874e-35, -2.1366e-34, 5.5972e-34, + 5.1575e-03, 8.8501e-04, 1.3161e-04, 4.2725e-04, 2.3460e-04, + 3.5889e-02, 1.1875e+00, 3.1836e-01, 4.7070e-01, 1.3438e+00, + 2.7539e-01, 3.0273e-01, 8.3923e-04, -1.8612e-30, 1.5799e-34, + -3.1296e-33, 1.0633e-04, 4.0627e-04, 4.3869e-05, 4.4678e-02, + 3.8528e-04, 3.6864e-35, -4.5515e-35, -1.3993e-34, -1.0472e-33, + 3.0632e-24, 2.0222e-33, 4.6094e-01, 9.4531e-01, 1.2344e+00, + 1.4844e+00, 2.3438e-01, 8.4375e-01, -6.0425e-03, -7.6294e-05, + -2.0703e-01, 3.1090e-04, -6.5000e-34, -3.7014e-34, -1.4264e-33, + 1.1587e-04, 3.5931e-24, 4.6945e-34, -1.0006e-34, -3.7616e-35, + -1.0382e-34, 3.2162e-35, -4.7396e-35], dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.self_attn.v_proj.weight': tensor([[ 0.0028, -0.0034, 0.0063, ..., -0.0203, -0.0049, 0.0070], + [ 0.0029, -0.0100, 0.0068, ..., 0.0069, -0.0041, -0.0063], + [ 0.0052, -0.0027, -0.0053, ..., -0.0231, -0.0046, -0.0111], + ..., + [ 0.0065, 0.0164, 0.0194, ..., -0.0010, 0.0004, -0.0175], + [-0.0047, -0.0121, 0.0077, ..., 0.0072, -0.0036, 0.0110], + [ 0.0112, -0.0040, -0.0160, ..., 0.0114, 0.0090, 0.0106]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.output_proj.weight': tensor([[-0.0054, -0.0135, -0.0113, ..., 0.0189, -0.0115, 0.0073], + [ 0.0178, -0.0056, -0.0044, ..., -0.0254, -0.0120, -0.0076], + [-0.0027, 0.0036, 0.0038, ..., -0.0048, 0.0053, 0.0042], + ..., + [ 0.0027, -0.0137, -0.0098, ..., -0.0210, -0.0160, -0.0188], + [ 0.0042, -0.0032, -0.0053, ..., -0.0211, 0.0063, -0.0063], + [ 0.0143, 0.0056, -0.0114, ..., -0.0033, 0.0105, -0.0081]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_proj.weight': tensor([[-0.0164, 0.0239, 0.0057, ..., 0.0036, -0.0100, 0.0140], + [ 0.0244, -0.0101, -0.0052, ..., -0.0085, 0.0110, -0.0208], + [ 0.0104, 0.0075, -0.0023, ..., -0.0057, 0.0073, -0.0005], + ..., + [ 0.0092, -0.0036, 0.0033, ..., 0.0065, -0.0099, 0.0070], + [-0.0061, -0.0043, -0.0024, ..., -0.0044, -0.0095, 0.0100], + [ 0.0219, 0.0141, -0.0095, ..., -0.0045, 0.0086, -0.0172]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_norm.weight': tensor([ 2.3682e-02, 4.8047e-01, 1.7822e-02, 8.7891e-03, 6.7749e-03, + 2.0606e-12, 9.5367e-06, 5.5469e-01, -1.8332e-12, 4.4727e-01, + 1.8799e-02, 1.9287e-02, 7.0496e-03, 2.2095e-02, 1.7944e-02, + 1.7456e-02, 6.0425e-03, 1.8215e-04, -3.6001e-05, 2.9297e-02, + 2.6562e-01, 2.2461e-02, 6.5231e-04, 1.8433e-02, 1.0757e-03, + 2.6131e-04, -1.7548e-04, 3.0884e-02, 6.5327e-05, 4.2969e-01, + -1.1969e-04, -4.2915e-05, 9.6436e-03, 1.4114e-04, 7.6904e-03, + 1.0376e-02, 2.0146e-05, 4.5703e-01, 1.5259e-02, 2.7771e-03, + 3.6377e-02, 1.8921e-02, 5.9204e-03, 2.8133e-05, 5.4321e-03, + 7.6294e-03, 1.8616e-03, 1.7578e-02, 2.8076e-02, 1.4893e-02, + 4.9438e-03, 4.4336e-01, 1.0071e-02, 1.5869e-02, 3.1494e-02, + 1.8692e-04, 6.8359e-02, 1.6479e-02, 1.5869e-02, 2.5391e-02, + 4.0770e-05, 1.1475e-02, 5.7373e-03, -4.8876e-05, 1.2756e-02, + 2.5482e-03, -1.8597e-05, -2.7776e-05, 8.9722e-03, 4.7852e-02, + -5.2387e-08, 6.1951e-03, 4.6143e-02, 3.2471e-02, 4.3640e-03, + 8.3008e-03, 2.9755e-03, 1.7578e-02, 2.1973e-02, 1.4648e-02, + -2.2793e-04, 1.1963e-02, 3.2471e-02, -2.8253e-05, 2.9419e-02, + 1.2360e-03, 4.3030e-03, 3.0518e-02, 1.1658e-02, 5.5847e-03, + 1.3306e-02, 5.9891e-04, 1.1292e-02, 2.1729e-02, 1.3306e-02, + -1.0395e-04, 1.9287e-02, 3.3203e-02, 1.5991e-02, 4.1748e-02, + 2.5513e-02, 2.3438e-02, 1.6357e-02, 1.0443e-04, -2.0373e-10, + 1.3733e-02, 4.6158e-04, 2.2461e-02, 1.6479e-02, 1.9531e-02, + 1.3184e-02, 5.8594e-03, 2.1118e-02, 5.0781e-01, 1.1063e-03, + 1.1353e-02, 7.8735e-03, 1.2512e-02, 5.1575e-03, 5.2002e-02, + 2.3499e-03, 2.3651e-03, -1.4210e-04, 1.2451e-02, 6.4087e-03, + 1.3367e-02, 2.0264e-02, 2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_proj.weight': tensor([[-0.0010, -0.0033, 0.0038, ..., -0.0006, 0.0034, 0.0026], + [ 0.0116, -0.0035, -0.0040, ..., 0.0050, 0.0197, 0.0012], + [ 0.0023, 0.0036, 0.0016, ..., -0.0006, -0.0019, 0.0029], + ..., + [-0.0017, -0.0049, -0.0036, ..., -0.0004, -0.0015, 0.0031], + [-0.0016, -0.0053, -0.0125, ..., 0.0029, 0.0045, 0.0010], + [ 0.0078, -0.0004, -0.0020, ..., -0.0027, 0.0015, -0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_norm.weight': tensor([ 2.4658e-02, 4.8047e-01, 1.7578e-02, 8.8501e-03, 6.7749e-03, + 2.1885e-12, 9.4771e-06, 5.5469e-01, 1.8474e-12, 4.4727e-01, + 1.9409e-02, 1.9165e-02, 7.0190e-03, 2.1973e-02, 1.7822e-02, + 1.7334e-02, 6.2866e-03, 1.7929e-04, -3.7432e-05, 2.9175e-02, + 2.6562e-01, 2.1973e-02, 6.5613e-04, 1.8555e-02, 1.0986e-03, + 2.6321e-04, -1.7357e-04, 3.0640e-02, 6.5327e-05, 4.2969e-01, + -1.1921e-04, -4.4823e-05, 9.4604e-03, 1.4019e-04, 7.6904e-03, + 1.0315e-02, 2.0504e-05, 4.5703e-01, 1.5259e-02, 2.6550e-03, + 3.5400e-02, 1.8921e-02, 5.5847e-03, 2.5511e-05, 5.3406e-03, + 7.6904e-03, 1.8768e-03, 1.7334e-02, 2.8198e-02, 1.5015e-02, + 4.8828e-03, 4.4336e-01, 1.0132e-02, 1.5747e-02, 3.1250e-02, + 1.8597e-04, 6.8848e-02, 1.6235e-02, 1.5869e-02, 2.5513e-02, + 4.1485e-05, 1.1536e-02, 5.7373e-03, -5.0783e-05, 1.2756e-02, + 2.4872e-03, -1.9431e-05, -2.7418e-05, 9.0942e-03, 4.7852e-02, + -1.1316e-07, 6.3477e-03, 4.6631e-02, 3.2471e-02, 4.4250e-03, + 8.6670e-03, 2.9297e-03, 1.7090e-02, 2.1973e-02, 1.4771e-02, + -2.2316e-04, 1.1780e-02, 3.1982e-02, -2.7776e-05, 2.9663e-02, + 1.0910e-03, 4.2725e-03, 3.0273e-02, 1.1963e-02, 5.7373e-03, + 1.3794e-02, 6.2561e-04, 1.1292e-02, 2.1606e-02, 1.3123e-02, + -1.0586e-04, 1.8677e-02, 3.2959e-02, 1.5869e-02, 4.1016e-02, + 2.6367e-02, 2.3560e-02, 1.6602e-02, 1.0347e-04, 2.0373e-10, + 1.3672e-02, 4.5013e-04, 2.2217e-02, 1.6602e-02, 1.9287e-02, + 1.3184e-02, 5.5847e-03, 2.0996e-02, 5.0781e-01, 1.0986e-03, + 1.1292e-02, 7.8735e-03, 1.2329e-02, 5.3711e-03, 5.2002e-02, + 2.3956e-03, 2.3956e-03, -1.4210e-04, 1.2634e-02, 6.3477e-03, + 1.3245e-02, 2.0264e-02, 2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.v_proj.weight': tensor([[-2.2583e-02, 2.2736e-03, 2.2430e-03, ..., -8.6784e-05, + -1.5442e-02, -2.7832e-02], + [-4.1199e-03, 1.0803e-02, 1.0376e-02, ..., 7.7820e-03, + 6.6833e-03, 7.7209e-03], + [-3.9307e-02, -5.0049e-03, -7.4158e-03, ..., -9.3994e-03, + -2.7710e-02, -1.2451e-02], + ..., + [-4.1748e-02, 7.0190e-04, -1.5625e-02, ..., 1.8158e-03, + 2.1729e-02, 9.3994e-03], + [ 6.5308e-03, 6.1646e-03, -4.4556e-03, ..., 3.6469e-03, + 8.6670e-03, 8.1787e-03], + [ 1.6235e-02, 4.6692e-03, -3.8757e-03, ..., 2.0294e-03, + 1.1353e-02, -4.6015e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.output_proj.weight': tensor([[ 0.0071, 0.0040, 0.0148, ..., 0.0079, 0.0043, -0.0017], + [-0.0080, -0.0045, -0.0041, ..., -0.0087, 0.0038, 0.0112], + [-0.0005, 0.0038, -0.0038, ..., 0.0015, -0.0027, -0.0022], + ..., + [ 0.0054, 0.0052, 0.0035, ..., -0.0064, -0.0051, 0.0078], + [ 0.0128, 0.0302, -0.0167, ..., -0.0222, 0.0045, 0.0425], + [-0.0012, -0.0087, -0.0229, ..., -0.0104, 0.0020, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_img.weight': tensor([[ 0.0079, 0.0020, 0.0430, ..., -0.0327, 0.0054, 0.0148], + [-0.0278, 0.0280, 0.0249, ..., -0.0068, -0.0378, 0.0271], + [ 0.0139, -0.0116, 0.0312, ..., -0.0075, 0.0253, -0.0110], + ..., + [ 0.0245, -0.0150, 0.0033, ..., -0.0452, 0.0216, -0.0209], + [-0.0081, 0.0073, -0.0471, ..., -0.0109, 0.0157, -0.0088], + [ 0.0078, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0038, -0.0123, -0.0115], + [ 0.0135, 0.0061, -0.0188, ..., 0.0157, 0.0082, -0.0042], + [-0.0306, -0.0028, -0.0272, ..., 0.0101, 0.0337, 0.0039], + ..., + [-0.0040, -0.0040, 0.0047, ..., 0.0209, -0.0103, 0.0073], + [ 0.0193, 0.0210, 0.0123, ..., -0.0145, 0.0117, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0110, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img.weight': tensor([[ 0.0058, -0.0056, 0.0173, ..., 0.0083, 0.0356, 0.0217], + [-0.0498, -0.0051, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0108, -0.0044, -0.0139, ..., 0.0039, 0.0269, 0.0077], + ..., + [-0.0043, 0.0125, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0077, -0.0162, ..., -0.0054, -0.0075, -0.0121], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0138, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.mlp.layer1.weight': tensor([[-0.0176, 0.0156, -0.0038, ..., -0.0092, -0.0087, -0.0221], + [-0.0026, -0.0106, 0.0096, ..., -0.0039, -0.0060, -0.0166], + [-0.0065, 0.0032, 0.0022, ..., -0.0078, -0.0032, -0.0050], + ..., + [ 0.0251, 0.0005, -0.0049, ..., 0.0010, 0.0085, 0.0071], + [-0.0251, 0.0047, -0.0145, ..., 0.0160, -0.0053, 0.0062], + [ 0.0092, -0.0067, -0.0008, ..., -0.0135, 0.0120, -0.0248]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.mlp.layer2.weight': tensor([[-0.0166, -0.0135, 0.0115, ..., -0.0025, 0.0060, -0.0061], + [-0.0049, 0.0071, 0.0084, ..., -0.0048, 0.0110, 0.0041], + [-0.0077, -0.0070, 0.0029, ..., -0.0098, 0.0082, -0.0077], + ..., + [-0.0027, 0.0090, 0.0045, ..., 0.0143, -0.0068, 0.0093], + [ 0.0049, 0.0103, 0.0046, ..., -0.0059, -0.0005, 0.0159], + [ 0.0105, -0.0038, -0.0099, ..., 0.0050, -0.0029, -0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0447, -0.0435, -0.0430, ..., 0.0050, 0.0053, 0.0053], + [ 0.0408, 0.0396, 0.0381, ..., -0.0037, -0.0036, -0.0037], + [-0.0454, -0.0437, -0.0432, ..., 0.0051, 0.0052, 0.0051], + ..., + [-0.0498, -0.0488, -0.0476, ..., 0.0035, 0.0036, 0.0037], + [-0.0439, -0.0432, -0.0425, ..., 0.0037, 0.0040, 0.0040], + [-0.0310, -0.0295, -0.0284, ..., 0.0011, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0058, 0.0006, 0.0046, ..., 0.0025, 0.0046, -0.0006], + [ 0.0018, -0.0023, -0.0004, ..., 0.0001, 0.0035, 0.0018], + [-0.0065, -0.0017, -0.0055, ..., -0.0077, -0.0070, 0.0017], + ..., + [-0.0293, 0.0015, -0.0344, ..., -0.0378, -0.0361, -0.0017], + [-0.0250, -0.0064, -0.0226, ..., -0.0249, -0.0222, 0.0072], + [ 0.0065, -0.0074, 0.0048, ..., 0.0065, -0.0005, 0.0062]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_cross_attn.1.weight': tensor([[ 6.6406e-02, 6.4941e-02, 6.3965e-02, ..., 4.8161e-05, + 1.6093e-05, 5.9843e-05], + [ 4.4556e-03, 4.4250e-03, 4.4250e-03, ..., -3.4332e-05, + -1.1063e-04, -2.2602e-04], + [-2.2827e-02, -2.1973e-02, -2.0996e-02, ..., -6.8283e-04, + -6.8283e-04, -6.9427e-04], + ..., + [-2.7344e-02, -2.6978e-02, -2.6245e-02, ..., -4.1580e-04, + -3.7575e-04, -3.3760e-04], + [-2.5391e-02, -2.4902e-02, -2.4170e-02, ..., -2.0599e-04, + -1.6022e-04, -2.1267e-04], + [ 5.8838e-02, 5.7617e-02, 5.6885e-02, ..., -1.4591e-04, + -2.6822e-05, -1.7166e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0094, -0.0076, -0.0112, ..., -0.0010, 0.0064, -0.0079], + [ 0.0042, 0.0039, 0.0047, ..., -0.0049, -0.0038, 0.0056], + [ 0.0130, -0.0038, -0.0038, ..., 0.0047, -0.0058, 0.0097], + ..., + [ 0.0679, 0.0107, -0.0051, ..., 0.0074, -0.0149, 0.0483], + [ 0.0288, -0.0056, -0.0077, ..., -0.0071, -0.0209, 0.0310], + [-0.0078, 0.0099, -0.0190, ..., 0.0144, 0.0126, -0.0093]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_mlp.1.weight': tensor([[-0.0043, -0.0045, -0.0044, ..., 0.0003, 0.0003, 0.0003], + [ 0.0659, 0.0640, 0.0635, ..., -0.0038, -0.0039, -0.0038], + [ 0.0415, 0.0408, 0.0403, ..., -0.0039, -0.0039, -0.0039], + ..., + [ 0.0320, 0.0309, 0.0306, ..., -0.0021, -0.0022, -0.0021], + [-0.0124, -0.0114, -0.0105, ..., 0.0014, 0.0014, 0.0013], + [ 0.0254, 0.0253, 0.0253, ..., -0.0025, -0.0026, -0.0026]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_mlp.2.weight': tensor([[-0.0028, 0.0018, 0.0028, ..., 0.0026, 0.0021, 0.0024], + [-0.0002, 0.0020, 0.0021, ..., 0.0019, -0.0001, 0.0010], + [ 0.0019, 0.0001, -0.0037, ..., -0.0031, -0.0019, -0.0029], + ..., + [-0.0029, 0.0013, -0.0189, ..., -0.0081, 0.0028, -0.0057], + [ 0.0151, -0.0075, -0.0244, ..., -0.0165, -0.0144, -0.0148], + [ 0.0172, -0.0195, -0.0133, ..., -0.0166, -0.0178, -0.0162]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.after_proj.weight': tensor([[ 0.0104, -0.0075, 0.0262, ..., -0.0078, 0.0193, -0.0015], + [-0.0023, -0.0077, 0.0088, ..., 0.0153, 0.0086, -0.0063], + [ 0.0081, 0.0028, -0.0168, ..., 0.0070, 0.0148, -0.0087], + ..., + [-0.0050, 0.0074, 0.0249, ..., -0.0206, -0.0128, -0.0078], + [-0.0036, 0.0177, 0.0115, ..., -0.0117, 0.0003, 0.0161], + [-0.0123, -0.0178, -0.0145, ..., 0.0075, 0.0110, -0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.after_proj.bias': tensor([-4.2419e-03, 5.1880e-03, 4.8523e-03, ..., 6.7749e-03, + 7.0572e-05, -1.3885e-03], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_proj.weight': tensor([[ 0.0249, -0.0265, -0.0237, ..., 0.0074, -0.0041, 0.0330], + [ 0.0048, 0.0044, -0.0016, ..., -0.0142, 0.0068, -0.0049], + [ 0.0253, 0.0031, -0.0038, ..., 0.0045, -0.0076, 0.0183], + ..., + [-0.0087, -0.0070, 0.0021, ..., -0.0044, -0.0036, -0.0004], + [ 0.0315, 0.0259, -0.0068, ..., -0.0079, -0.0048, 0.0042], + [-0.0124, -0.0121, 0.0179, ..., -0.0028, 0.0063, -0.0244]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_norm.weight': tensor([ 1.3750e+00, 5.2344e-01, 6.6406e-01, 4.7852e-01, 1.3359e+00, + 1.4893e-02, 9.6436e-03, 1.9789e-05, 2.1057e-03, 1.5137e-02, + 3.3677e-06, -7.6252e-09, 1.1406e+00, -2.5988e-05, -3.5508e-18, + -4.7119e-02, 1.1597e-02, 2.6489e-02, 2.6703e-04, 1.0010e-02, + 1.5527e-01, -3.7909e-05, 9.6875e-01, 1.5938e+00, 7.1094e-01, + 1.1016e+00, 7.1094e-01, 1.9141e-01, 2.3682e-02, -1.4099e-02, + 7.1106e-03, 3.1090e-04, 7.5989e-03, -3.0212e-03, -1.0071e-03, + 2.9663e-02, 1.3965e-01, -2.5272e-05, 5.9814e-03, 2.1973e-03, + -7.0930e-06, -6.2883e-06, -1.7956e-06, 1.5625e+00, 6.4844e-01, + 1.5000e+00, 1.0625e+00, 4.8633e-01, -3.2227e-02, -3.9673e-03, + -9.9182e-04, 2.7222e-02, 1.3184e-01, -3.9101e-05, 6.4373e-05, + -4.3213e-06, -2.4033e-04, -5.0659e-03, -5.5176e-02, -3.1832e-10, + 3.2471e-02, -1.6809e-05, -2.2461e-02, -2.8320e-02, 8.9844e-01, + 1.2422e+00, 1.3438e+00, 1.3828e+00, 6.1328e-01, -2.0508e-02, + 9.5215e-03, 8.8215e-05, 1.2024e-02, 1.2283e-03, 9.0525e-07, + 1.7695e-07, 6.6895e-02, 9.3460e-05, 9.1903e-20, 1.0000e+00, + 1.1797e+00, 2.8564e-02, 2.2507e-04, 3.1494e-02, 1.1475e-01, + 1.1475e-02, 1.6719e+00, 9.8438e-01, 1.4609e+00, 1.5391e+00, + 2.9102e-01, 6.1328e-01, 5.2490e-02, -8.4229e-03, -1.1719e-02, + 1.7357e-04, -3.0396e-02, -1.8677e-02, -3.6316e-03, 4.1260e-02, + 1.3281e-01, -4.4584e-05, 1.1875e+00, -3.6478e-05, -9.1195e-06, + -9.1553e-05, -5.9372e-08, 1.1875e+00, 1.6328e+00, 1.5469e+00, + 1.5703e+00, 8.1250e-01, 1.9775e-02, -2.6001e-02, 3.2959e-03, + 3.4424e-02, 1.3770e-01, 1.4544e-05, 4.1723e-05, -3.1441e-06, + -4.1723e-06, -6.1512e-05, 1.2266e+00, -8.3674e-11, 3.0640e-02, + -5.6028e-06, 9.6484e-01, 1.2188e+00], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.self_attn.k_proj.weight': tensor([[-0.0034, 0.0308, 0.0071, ..., 0.0121, 0.0184, 0.0039], + [ 0.0065, 0.0037, 0.0056, ..., -0.0039, 0.0052, 0.0047], + [-0.0232, 0.0197, 0.0015, ..., 0.0042, 0.0145, 0.0150], + ..., + [ 0.0044, -0.0096, -0.0056, ..., 0.0079, 0.0087, 0.0111], + [ 0.0140, 0.0077, 0.0138, ..., -0.0079, -0.0084, 0.0176], + [ 0.0022, -0.0189, 0.0095, ..., -0.0082, 0.0088, 0.0299]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.k_norm.weight': tensor([ 1.4141e+00, 7.5781e-01, 4.7070e-01, 4.7461e-01, 1.3438e+00, + 1.5991e-02, 1.3977e-02, 8.8692e-05, 7.1716e-03, 1.5076e-02, + 3.2634e-06, -4.3656e-09, 1.1406e+00, 4.3869e-05, -3.5779e-18, + -3.7537e-03, -1.1206e-04, 9.2773e-02, 3.2196e-03, -3.2196e-03, + 4.0283e-03, -3.9339e-05, 1.0625e+00, 1.4609e+00, 4.9609e-01, + 1.2266e+00, 7.8125e-01, 2.9102e-01, 1.2451e-02, 5.8289e-03, + 1.0254e-02, 3.3379e-04, 8.3008e-03, 4.3213e-02, 1.8539e-03, + 9.0332e-03, 2.7710e-02, 1.7643e-05, 2.1729e-02, 2.1973e-03, + -8.8811e-06, -2.3395e-06, 1.6317e-06, 1.7188e+00, 1.1875e+00, + 1.5156e+00, 1.2734e+00, 4.8828e-01, -1.6357e-02, 6.4453e-02, + 3.1281e-03, -9.9487e-03, 4.7913e-03, -3.2902e-05, -1.7047e-05, + 3.9935e-06, -2.5177e-04, -5.0659e-03, 3.0640e-02, 2.7831e-10, + -3.9673e-03, -1.6212e-05, -3.0029e-02, -2.9907e-02, 6.8359e-01, + 1.1406e+00, 1.3750e+00, 1.3594e+00, 4.8242e-01, -1.9165e-02, + 6.1035e-03, -4.5300e-06, 3.5553e-03, -9.6893e-04, 1.1921e-06, + 1.8068e-07, 5.3406e-03, 5.8889e-05, 6.7424e-19, 1.0000e+00, + 1.1797e+00, 8.1055e-02, -4.8828e-04, -3.8300e-03, -3.6316e-03, + 1.1597e-02, 1.5859e+00, 1.2266e+00, 1.5703e+00, 1.5703e+00, + 8.7402e-02, 6.0547e-01, 2.3193e-02, 1.3977e-02, -7.9346e-03, + 1.0395e-04, 3.1738e-02, 6.0303e-02, 7.0801e-03, -1.0254e-02, + -2.5879e-02, -5.1022e-05, 1.1875e+00, 3.6478e-05, -2.4199e-05, + -9.1076e-05, -6.9384e-08, 1.0469e+00, 1.4766e+00, 1.4609e+00, + 1.6797e+00, 8.2031e-01, -2.5024e-02, 9.8633e-02, 3.6163e-03, + 1.9165e-02, -1.6602e-02, -2.7657e-05, 3.2425e-05, -2.9951e-06, + -2.0266e-06, 7.4387e-05, 1.2266e+00, -2.8081e-11, -1.0742e-02, + 5.3346e-06, 9.6484e-01, 1.2188e+00], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.self_attn.v_proj.weight': tensor([[-0.0051, 0.0036, -0.0222, ..., -0.0315, 0.0049, 0.0051], + [ 0.0030, 0.0017, -0.0138, ..., 0.0074, 0.0099, 0.0056], + [-0.0173, -0.0101, 0.0111, ..., 0.0151, -0.0081, -0.0086], + ..., + [ 0.0072, 0.0109, -0.0194, ..., -0.0356, 0.0075, 0.0027], + [ 0.0049, 0.0150, -0.0041, ..., 0.0060, 0.0060, -0.0049], + [ 0.0119, -0.0081, -0.0173, ..., 0.0171, -0.0041, 0.0049]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.output_proj.weight': tensor([[ 0.0085, 0.0261, -0.0267, ..., 0.0055, -0.0234, 0.0265], + [-0.0041, -0.0102, -0.0131, ..., 0.0220, 0.0256, -0.0087], + [ 0.0023, 0.0040, -0.0027, ..., -0.0017, 0.0041, 0.0084], + ..., + [ 0.0037, 0.0060, -0.0251, ..., 0.0096, 0.0140, -0.0101], + [-0.0144, 0.0079, 0.0062, ..., 0.0143, 0.0064, -0.0052], + [ 0.0088, -0.0034, -0.0133, ..., 0.0095, 0.0078, 0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_proj.weight': tensor([[-0.0112, 0.0037, -0.0019, ..., -0.0033, -0.0048, -0.0020], + [ 0.0121, 0.0047, 0.0027, ..., -0.0041, -0.0039, 0.0050], + [-0.0042, -0.0073, -0.0038, ..., -0.0074, 0.0123, 0.0027], + ..., + [-0.0063, -0.0048, -0.0037, ..., 0.0152, -0.0041, -0.0047], + [-0.0011, -0.0074, 0.0025, ..., 0.0134, -0.0055, -0.0047], + [ 0.0025, 0.0034, 0.0023, ..., -0.0037, 0.0043, 0.0037]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_norm.weight': tensor([-6.5613e-04, 1.3184e-01, 1.4526e-02, 6.0547e-01, 2.2339e-02, + 1.2634e-02, 8.5449e-03, 3.2806e-03, 7.0190e-03, 1.1279e-01, + 1.8677e-02, 5.0049e-02, 2.6367e-01, 1.3977e-02, 1.7090e-01, + 1.3245e-02, 5.3467e-02, 3.1494e-02, 1.6937e-03, 1.7578e-02, + 8.6060e-03, 1.8262e-01, 2.7344e-01, 1.8555e-01, 2.0752e-02, + 1.6016e-01, 5.5908e-02, 9.7656e-03, 4.0234e-01, 6.2866e-03, + 1.1292e-02, 5.0781e-01, 5.5664e-02, 3.3936e-02, 1.6113e-02, + 1.3123e-02, 6.7871e-02, 4.5508e-01, 5.3906e-01, 1.4954e-02, + 3.8910e-03, 1.0449e-01, 5.2490e-03, 5.3906e-01, 1.8188e-02, + 4.3213e-02, 1.0437e-02, 3.9258e-01, 7.2327e-03, 1.7944e-02, + 6.8359e-02, 1.8799e-02, 1.0437e-02, 5.0391e-01, -8.5831e-05, + 1.9897e-02, 5.5664e-02, 7.7148e-02, 9.0942e-03, 1.0132e-02, + 5.7983e-03, 9.6436e-03, 1.4099e-02, 3.3008e-01, 2.0264e-02, + 2.8534e-03, 2.6001e-02, 5.2979e-02, 6.2988e-02, 1.6235e-02, + 6.1951e-03, 6.8359e-01, -3.4094e-05, 6.1328e-01, 9.0332e-03, + 1.5198e-02, 2.2217e-02, 2.6953e-01, 3.2471e-02, 9.1171e-04, + 1.0645e-01, 9.7656e-03, 4.0771e-02, 1.1169e-02, 2.1362e-02, + 7.7209e-03, 2.9602e-03, 6.9336e-02, 1.9434e-01, 8.9111e-03, + 9.3994e-03, 1.5625e-02, 4.6875e-02, 2.8320e-02, 2.3193e-02, + 2.3560e-02, 1.4709e-02, 1.6846e-02, 6.2012e-02, 1.9287e-02, + 3.1250e-02, 7.2632e-03, 1.0803e-02, 5.4932e-02, 6.5234e-01, + 6.8848e-02, 2.6550e-03, 4.4531e-01, 1.0925e-02, 1.8799e-02, + 1.6479e-02, 1.5015e-02, 8.4043e-06, 8.9844e-02, 4.2969e-01, + 2.4414e-02, 1.7334e-02, 6.2109e-01, 1.1914e-01, 4.0039e-02, + 1.2146e-02, 7.1875e-01, 1.2891e-01, 1.8799e-02, 1.0925e-02, + 4.4250e-03, 2.0752e-02, 1.3046e-03], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_proj.weight': tensor([[-0.0046, -0.0034, 0.0103, ..., 0.0059, -0.0056, 0.0032], + [-0.0042, 0.0003, -0.0034, ..., 0.0005, -0.0081, 0.0029], + [-0.0002, -0.0016, -0.0085, ..., 0.0123, -0.0168, -0.0072], + ..., + [ 0.0040, 0.0046, -0.0043, ..., 0.0052, 0.0020, -0.0061], + [ 0.0093, 0.0029, -0.0031, ..., -0.0035, 0.0093, 0.0023], + [-0.0056, 0.0007, -0.0092, ..., -0.0081, -0.0060, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_norm.weight': tensor([ 6.8665e-04, 1.3184e-01, 1.4038e-02, 6.0156e-01, 2.1851e-02, + 1.2573e-02, 8.7280e-03, 3.2654e-03, 7.4158e-03, 1.1279e-01, + 1.8677e-02, 4.8584e-02, 2.6367e-01, 1.3855e-02, 1.7090e-01, + 1.3916e-02, 5.2734e-02, 3.1738e-02, 1.6327e-03, 1.7456e-02, + 8.9111e-03, 1.8359e-01, 2.7539e-01, 1.8555e-01, 2.0508e-02, + 1.6211e-01, 5.4443e-02, 7.9956e-03, 4.0430e-01, 6.2561e-03, + 1.1536e-02, 5.0781e-01, 5.5664e-02, 3.3936e-02, 1.6479e-02, + 1.2451e-02, 6.8359e-02, 4.5508e-01, 5.3906e-01, 1.4954e-02, + 3.8910e-03, 1.0449e-01, 5.2185e-03, 5.3906e-01, 1.8433e-02, + 4.3213e-02, 1.0803e-02, 3.9258e-01, 7.3853e-03, 1.7822e-02, + 6.8359e-02, 1.9043e-02, 1.0376e-02, 5.0391e-01, 1.2589e-04, + 1.9897e-02, 5.5664e-02, 7.7148e-02, 9.0942e-03, 8.3008e-03, + 5.7373e-03, 9.6436e-03, 1.4404e-02, 3.3008e-01, 1.9775e-02, + 2.9755e-03, 2.5513e-02, 5.2979e-02, 6.2988e-02, 1.6235e-02, + 6.5918e-03, 6.7969e-01, -3.3855e-05, 6.1328e-01, 9.2163e-03, + -1.4771e-02, 2.2217e-02, 2.6953e-01, 3.3203e-02, 9.1934e-04, + 1.0693e-01, 9.6436e-03, 4.0039e-02, 1.0986e-02, 2.0874e-02, + 7.7820e-03, 3.9062e-03, 6.9824e-02, 1.9434e-01, 8.9111e-03, + 9.5825e-03, 1.4709e-02, 4.7119e-02, 2.8442e-02, 2.4048e-02, + 2.3682e-02, 1.4709e-02, 1.6968e-02, 6.2012e-02, 1.9165e-02, + 3.1250e-02, 7.4768e-03, 1.1230e-02, 5.5420e-02, 6.4844e-01, + 6.8848e-02, 2.6855e-03, 4.4727e-01, 1.0620e-02, 1.8677e-02, + 1.6968e-02, 1.5015e-02, 1.1623e-05, 8.9355e-02, 4.2969e-01, + 2.4170e-02, 1.7212e-02, 6.2109e-01, 1.1914e-01, 4.0283e-02, + 1.2085e-02, 7.2266e-01, 1.2793e-01, 1.7822e-02, 1.0681e-02, + 4.3640e-03, 2.0630e-02, 7.3624e-04], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.v_proj.weight': tensor([[ 0.0170, -0.0073, -0.0155, ..., 0.0033, -0.0150, -0.0077], + [-0.0186, -0.0008, -0.0024, ..., -0.0024, -0.0233, 0.0194], + [-0.0188, -0.0067, -0.0031, ..., -0.0038, 0.0283, 0.0312], + ..., + [-0.0130, -0.0038, -0.0086, ..., -0.0001, 0.0154, 0.0029], + [ 0.0229, 0.0045, 0.0222, ..., 0.0052, 0.0056, 0.0077], + [ 0.0139, 0.0012, 0.0033, ..., 0.0089, -0.0164, -0.0210]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0003, 0.0027, ..., -0.0012, 0.0021, 0.0156], + [ 0.0109, -0.0014, 0.0027, ..., 0.0042, 0.0002, -0.0014], + [ 0.0042, 0.0008, -0.0021, ..., -0.0047, -0.0039, -0.0054], + ..., + [-0.0178, -0.0018, 0.0037, ..., 0.0044, -0.0113, 0.0034], + [ 0.0046, -0.0100, 0.0056, ..., -0.0310, 0.0078, -0.0223], + [ 0.0013, -0.0036, 0.0085, ..., 0.0297, -0.0044, 0.0217]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0064, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [-0.0015, 0.0150, -0.0159, ..., -0.0041, -0.0055, 0.0679], + ..., + [-0.0208, 0.0130, -0.0273, ..., 0.0308, -0.0072, -0.0068], + [-0.0337, -0.0025, 0.0054, ..., -0.0040, 0.0172, 0.0231], + [-0.0255, 0.0040, 0.0030, ..., 0.0408, 0.0430, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0031], + [ 0.0198, -0.0208, 0.0130, ..., 0.0089, -0.0337, -0.0055], + [ 0.0302, 0.0101, -0.0023, ..., -0.0304, 0.0199, 0.0024], + ..., + [ 0.0171, -0.0275, -0.0075, ..., 0.0231, 0.0182, -0.0039], + [ 0.0221, 0.0388, 0.0056, ..., 0.0042, -0.0488, -0.0361], + [-0.0005, 0.0120, -0.0051, ..., 0.0037, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img.weight': tensor([[ 0.0273, -0.0041, 0.0118, ..., 0.0012, 0.0347, 0.0615], + [-0.0258, 0.0042, -0.0197, ..., 0.0139, -0.0047, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0097, 0.0132, 0.0184, ..., 0.0110, 0.0087, 0.0515], + [-0.0006, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.mlp.layer1.weight': tensor([[ 0.0017, -0.0165, 0.0046, ..., 0.0017, -0.0049, -0.0064], + [-0.0162, 0.0078, -0.0029, ..., 0.0124, 0.0065, -0.0203], + [-0.0122, -0.0107, 0.0084, ..., -0.0070, -0.0010, -0.0610], + ..., + [-0.0042, 0.0038, 0.0038, ..., 0.0156, 0.0085, -0.0038], + [-0.0093, 0.0140, -0.0031, ..., -0.0092, 0.0054, -0.0396], + [-0.0037, 0.0162, -0.0016, ..., 0.0114, 0.0172, -0.0091]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.mlp.layer2.weight': tensor([[ 0.0040, -0.0082, -0.0236, ..., 0.0062, -0.0112, -0.0124], + [-0.0309, -0.0120, -0.0055, ..., -0.0011, -0.0064, -0.0074], + [ 0.0049, -0.0071, -0.0030, ..., 0.0031, -0.0033, 0.0087], + ..., + [-0.0028, -0.0079, -0.0432, ..., -0.0063, -0.0098, -0.0110], + [-0.0104, 0.0050, 0.0120, ..., 0.0024, 0.0128, -0.0052], + [-0.0078, -0.0063, -0.0037, ..., -0.0096, 0.0097, -0.0388]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0396, 0.0398, 0.0381, ..., -0.0042, -0.0042, -0.0044], + [-0.0383, -0.0369, -0.0359, ..., 0.0043, 0.0048, 0.0047], + [-0.0400, -0.0393, -0.0383, ..., 0.0045, 0.0049, 0.0049], + ..., + [-0.0422, -0.0408, -0.0400, ..., 0.0022, 0.0022, 0.0024], + [-0.0415, -0.0408, -0.0398, ..., 0.0021, 0.0022, 0.0022], + [-0.0039, -0.0039, -0.0039, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_self_attn.2.weight': tensor([[ 3.8757e-03, -3.4637e-03, -3.5248e-03, ..., 2.0504e-05, + -5.6744e-05, 5.4016e-03], + [ 3.3722e-03, -4.0283e-03, -3.2654e-03, ..., -4.7302e-03, + -4.6997e-03, -1.7776e-03], + [ 4.8218e-03, -6.3171e-03, -5.7678e-03, ..., -3.7842e-03, + -3.8147e-03, 1.3123e-03], + ..., + [ 3.9062e-02, -4.1992e-02, -4.2480e-02, ..., -2.0630e-02, + -1.7334e-02, 1.6235e-02], + [ 3.9978e-03, -5.7983e-03, -5.8899e-03, ..., 3.8910e-03, + 3.8910e-03, 5.5847e-03], + [-1.4038e-02, 1.5564e-02, 1.4099e-02, ..., 1.7090e-02, + 1.6968e-02, 1.4526e-02]], dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9570e-02, 5.8838e-02, 5.7617e-02, ..., -1.1015e-04, + -3.9339e-05, -1.9169e-04], + [ 6.3965e-02, 6.2012e-02, 5.9326e-02, ..., 3.4904e-04, + 4.6158e-04, 4.9210e-04], + [-4.8340e-02, -4.6387e-02, -4.5654e-02, ..., -2.6464e-05, + 2.8014e-05, -8.0109e-05], + ..., + [ 3.1494e-02, 3.0518e-02, 2.9785e-02, ..., 1.1206e-04, + 4.5061e-05, 7.3433e-05], + [ 6.1035e-02, 5.9326e-02, 5.8594e-02, ..., 5.2643e-04, + 5.6076e-04, 5.5313e-04], + [-4.7852e-02, -4.6143e-02, -4.5166e-02, ..., -3.2234e-04, + -3.6430e-04, -3.2806e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0149, 0.0073, -0.0080, ..., 0.0018, -0.0047, -0.0041], + [-0.0178, 0.0121, 0.0144, ..., -0.0096, -0.0124, 0.0083], + [ 0.0011, 0.0030, -0.0011, ..., 0.0009, 0.0003, -0.0002], + ..., + [ 0.0260, 0.0184, -0.0149, ..., 0.0078, 0.0080, 0.0037], + [ 0.0075, 0.0334, -0.0057, ..., 0.0076, 0.0325, -0.0178], + [-0.0044, 0.0203, 0.0051, ..., -0.0114, -0.0110, 0.0240]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_mlp.1.weight': tensor([[-0.0986, -0.0967, -0.0942, ..., 0.0031, 0.0034, 0.0033], + [-0.0371, -0.0352, -0.0339, ..., 0.0008, 0.0008, 0.0007], + [ 0.0566, 0.0549, 0.0544, ..., -0.0017, -0.0017, -0.0017], + ..., + [ 0.0364, 0.0354, 0.0347, ..., -0.0018, -0.0018, -0.0016], + [ 0.0466, 0.0444, 0.0430, ..., -0.0010, -0.0009, -0.0010], + [-0.0518, -0.0505, -0.0496, ..., 0.0022, 0.0022, 0.0021]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_mlp.2.weight': tensor([[-0.0228, -0.0165, 0.0110, ..., -0.0018, 0.0181, -0.0115], + [ 0.0017, -0.0033, -0.0016, ..., -0.0016, 0.0030, 0.0016], + [-0.0005, -0.0007, -0.0008, ..., -0.0010, 0.0007, 0.0009], + ..., + [ 0.0454, -0.0031, -0.0184, ..., -0.0126, 0.0032, 0.0193], + [ 0.0459, -0.0288, -0.0371, ..., -0.0337, 0.0286, 0.0366], + [ 0.0128, -0.0366, -0.0251, ..., -0.0284, 0.0374, 0.0240]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.after_proj.weight': tensor([[-0.0054, -0.0084, -0.0073, ..., 0.0059, 0.0018, -0.0046], + [ 0.0066, 0.0057, 0.0067, ..., 0.0050, 0.0052, 0.0131], + [-0.0096, -0.0041, -0.0051, ..., -0.0023, -0.0037, 0.0110], + ..., + [-0.0015, -0.0067, -0.0140, ..., -0.0055, -0.0112, 0.0187], + [-0.0049, -0.0150, -0.0076, ..., 0.0105, -0.0118, -0.0096], + [-0.0036, 0.0160, 0.0104, ..., -0.0038, 0.0221, -0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.after_proj.bias': tensor([-3.4790e-03, 2.1820e-03, 1.1139e-03, ..., 4.6387e-03, + 9.3937e-05, -4.8637e-04], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_proj.weight': tensor([[-0.0110, 0.0153, 0.0214, ..., 0.0090, 0.0057, -0.0190], + [ 0.0203, -0.0014, 0.0074, ..., 0.0058, -0.0027, 0.0075], + [-0.0127, -0.0063, 0.0129, ..., -0.0188, -0.0093, -0.0134], + ..., + [ 0.0065, 0.0038, 0.0002, ..., 0.0057, 0.0030, 0.0051], + [ 0.0098, -0.0222, -0.0029, ..., -0.0398, -0.0059, -0.0120], + [-0.0065, -0.0048, -0.0052, ..., 0.0099, 0.0019, 0.0074]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_norm.weight': tensor([ 1.3672e+00, 1.8047e+00, 1.3828e+00, 1.5859e+00, 2.2852e-01, + 5.6396e-02, 6.5308e-03, 4.2915e-06, -1.7694e-33, 3.5477e-04, + 3.9062e-02, 1.2695e-01, 1.0620e-02, 3.2959e-02, -1.1719e-02, + 4.9744e-03, 6.3965e-02, 9.5703e-01, 8.8281e-01, 1.0376e-03, + 8.9169e-05, 1.2734e+00, 2.2188e+00, 1.9609e+00, 1.0469e+00, + 1.1094e+00, 1.5312e+00, 8.9264e-04, -1.6689e-04, -5.9128e-04, + 1.1292e-02, 1.7944e-02, 9.9487e-03, 1.0156e+00, 9.1553e-03, + 1.0156e+00, 7.3242e-04, -6.8665e-05, 7.4768e-03, -1.2024e-02, + 1.9670e-06, -5.9605e-06, 6.8665e-05, 2.3125e+00, 1.5000e+00, + 1.5781e+00, 1.6016e+00, 1.7578e+00, 1.1523e-01, 7.4768e-04, + 2.4048e-02, 2.4707e-01, 1.1172e+00, 1.3281e-01, 1.4038e-03, + 2.1973e-01, 3.0327e-04, 1.2305e-01, 6.1035e-05, -4.4861e-03, + -7.4463e-03, -5.6152e-03, -6.3477e-02, 1.1328e+00, 1.9609e+00, + 8.3984e-01, 1.1094e+00, 1.3281e+00, 7.7637e-02, -2.1484e-02, + 1.2573e-02, 2.5511e-05, -2.9491e-33, -1.1597e-03, 5.2795e-03, + 1.0559e-02, 1.4496e-03, -3.5156e-02, -2.8442e-02, 1.7944e-02, + 8.8867e-02, 3.6133e-02, 1.0254e-02, 6.1951e-03, -3.7994e-03, + 2.8687e-03, 1.0156e+00, 1.1953e+00, 1.6094e+00, 1.5703e+00, + 9.3750e-01, -7.4768e-03, 5.8365e-04, -2.9297e-03, 3.1281e-03, + -6.8665e-03, 2.1729e-02, 5.6152e-02, 1.7090e-02, 4.4678e-02, + -2.8229e-03, -1.3199e-03, 1.5234e-01, 4.2439e-05, 5.4550e-04, + -8.8811e-06, 2.8534e-03, 8.0469e-01, 1.7734e+00, 1.5547e+00, + 1.1016e+00, 1.2031e+00, 6.2988e-02, 6.5994e-04, 4.1992e-02, + 1.2188e+00, 1.1035e-01, 1.1641e+00, -2.0790e-04, 1.3794e-02, + 1.1475e-02, 5.6641e-02, 2.0313e-04, -2.2949e-02, -5.6028e-05, + 1.2695e-02, 1.0234e+00, -9.7656e-03], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.self_attn.k_proj.weight': tensor([[-0.0220, -0.0096, -0.0002, ..., -0.0080, 0.0030, -0.0188], + [ 0.0066, -0.0347, -0.0034, ..., 0.0116, 0.0139, 0.0144], + [-0.0121, 0.0195, -0.0061, ..., -0.0107, -0.0099, 0.0030], + ..., + [ 0.0010, 0.0106, -0.0084, ..., -0.0008, 0.0042, 0.0038], + [ 0.0222, -0.0081, 0.0102, ..., -0.0147, -0.0076, 0.0221], + [-0.0123, -0.0031, -0.0052, ..., 0.0056, 0.0062, 0.0053]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.k_norm.weight': tensor([ 9.6094e-01, 1.7266e+00, 1.3984e+00, 1.4844e+00, 2.3535e-01, + 6.2500e-02, -1.8845e-03, 1.1861e-05, -1.5167e-33, 2.5177e-04, + 3.9551e-02, 1.2793e-01, -1.1780e-02, 5.1025e-02, -1.1902e-02, + -4.7607e-03, 1.0925e-02, 9.5703e-01, 8.8281e-01, -9.4986e-04, + 2.2316e-04, 1.2734e+00, 1.8828e+00, 1.7734e+00, 4.3945e-02, + 5.3906e-01, 1.5469e+00, -6.9580e-03, -3.4714e-04, 4.4823e-04, + -1.1597e-02, 1.7822e-02, -6.1340e-03, 1.0078e+00, -7.6599e-03, + 1.0078e+00, 7.4768e-04, -1.0681e-04, -9.5215e-03, -1.2146e-02, + 2.9206e-06, -7.8082e-06, 6.4373e-05, 1.8672e+00, 1.0234e+00, + 1.5391e+00, 1.6641e+00, 1.7109e+00, 7.5684e-02, 8.0872e-04, + -3.9368e-03, -6.7444e-03, 1.1172e+00, 1.7188e-01, 1.3351e-03, + 2.1973e-01, -8.8501e-04, 1.2305e-01, -1.3173e-05, 7.1289e-02, + -7.4768e-03, 2.4414e-02, -7.1777e-02, 1.1328e+00, 2.0781e+00, + 1.2109e+00, 9.2969e-01, 1.3828e+00, -1.7944e-02, -1.0254e-02, + -1.6968e-02, -2.2888e-05, 4.3815e-33, 1.3275e-03, 1.7456e-02, + 4.6143e-02, 6.2561e-04, 4.9561e-02, -3.9795e-02, -1.7944e-02, + -5.6152e-02, 2.8809e-02, 1.3062e-02, 6.1340e-03, 3.9673e-03, + -1.1215e-03, 1.5234e+00, 1.5625e+00, 1.6953e+00, 1.6172e+00, + 5.4297e-01, 5.5542e-03, -4.3297e-04, -2.9907e-03, 9.4223e-04, + 7.1411e-03, -3.7384e-03, 5.9509e-03, -1.6235e-02, 8.3008e-03, + -2.8076e-03, -1.3275e-03, 1.5234e-01, 3.1233e-05, 5.6839e-04, + -1.0788e-05, 2.8534e-03, 1.6250e+00, 1.9844e+00, 1.5469e+00, + 5.5859e-01, 1.2969e+00, -4.3030e-03, 2.6512e-04, -3.1738e-02, + 1.2188e+00, 1.5442e-02, 1.1641e+00, -2.6512e-04, 1.6724e-02, + 1.1719e-02, 5.5908e-02, 2.0313e-04, 5.7617e-02, -1.0014e-04, + 2.2949e-02, 1.0234e+00, 3.2959e-03], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.self_attn.v_proj.weight': tensor([[ 0.0020, -0.0152, 0.0064, ..., 0.0461, -0.0211, 0.0038], + [ 0.0127, 0.0267, -0.0182, ..., 0.0046, 0.0138, -0.0134], + [-0.0003, 0.0297, -0.0025, ..., -0.0084, 0.0220, 0.0025], + ..., + [ 0.0086, 0.0102, -0.0201, ..., 0.0080, 0.0078, 0.0047], + [-0.0018, 0.0332, 0.0127, ..., 0.0145, 0.0266, -0.0056], + [-0.0086, -0.0042, 0.0227, ..., 0.0187, 0.0172, -0.0100]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.output_proj.weight': tensor([[ 0.0066, 0.0055, 0.0227, ..., 0.0034, 0.0201, -0.0080], + [ 0.0229, -0.0079, 0.0054, ..., -0.0186, 0.0041, -0.0114], + [ 0.0059, 0.0130, -0.0233, ..., -0.0077, -0.0072, 0.0134], + ..., + [-0.0129, -0.0464, -0.0078, ..., -0.0089, -0.0006, -0.0193], + [ 0.0032, 0.0084, -0.0201, ..., 0.0087, 0.0236, 0.0068], + [-0.0566, -0.0126, 0.0322, ..., 0.0216, -0.0221, -0.0166]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_proj.weight': tensor([[ 0.0098, -0.0079, 0.0138, ..., 0.0038, 0.0074, 0.0201], + [ 0.0078, -0.0058, -0.0052, ..., -0.0090, -0.0052, 0.0119], + [-0.0033, 0.0165, -0.0048, ..., 0.0044, -0.0076, -0.0115], + ..., + [ 0.0093, 0.0032, -0.0132, ..., 0.0160, 0.0046, -0.0068], + [-0.0110, -0.0153, -0.0032, ..., -0.0090, 0.0082, 0.0144], + [ 0.0095, -0.0047, -0.0076, ..., 0.0019, 0.0171, -0.0103]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_norm.weight': tensor([0.1064, 0.0159, 0.4277, 0.0698, 0.1680, 0.0361, 0.1484, 0.1357, 0.5938, + 0.0364, 0.0618, 0.0195, 0.0491, 0.0767, 0.0566, 0.1963, 0.4844, 0.0320, + 0.7461, 0.0352, 0.0437, 0.1689, 0.0178, 0.0801, 0.0481, 0.4082, 0.5508, + 0.2559, 0.0771, 0.0535, 0.0403, 0.2158, 0.6680, 0.2441, 0.0518, 0.0181, + 0.0105, 0.1099, 0.0156, 0.0476, 0.0493, 0.1289, 0.0182, 0.0400, 0.2041, + 0.0532, 0.3555, 0.0205, 0.0500, 0.0139, 0.0310, 0.3848, 0.5391, 0.3438, + 0.0903, 0.6484, 0.5977, 0.0400, 0.1235, 0.0322, 0.0854, 0.5586, 0.0102, + 0.5508, 0.1719, 0.3926, 0.0825, 0.0508, 0.0625, 0.4688, 0.0242, 0.5234, + 0.0214, 0.0356, 0.1338, 0.0874, 0.0287, 0.0889, 0.6367, 0.0352, 0.0889, + 0.0825, 0.0435, 0.0391, 0.4219, 0.0640, 0.0625, 0.0986, 0.0352, 0.1426, + 0.0459, 0.1758, 0.0537, 0.0588, 0.0295, 0.1221, 0.0131, 0.3613, 0.5820, + 0.0342, 0.0562, 0.2949, 0.0208, 0.1211, 0.0442, 0.2314, 0.0228, 0.6328, + 0.5859, 0.0327, 0.0120, 0.7695, 0.0192, 0.0137, 0.4941, 0.2070, 0.2754, + 0.7500, 0.0410, 0.0757, 0.0776, 0.5781, 0.6406, 0.0125, 0.0272, 0.2021, + 0.0249, 0.0205], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_proj.weight': tensor([[-8.7280e-03, 1.8158e-03, 7.0496e-03, ..., -5.6458e-04, + 2.1973e-02, -2.3346e-03], + [ 7.2327e-03, -6.7711e-05, -4.7302e-03, ..., -2.6398e-03, + -2.5330e-03, 1.0986e-02], + [-1.3123e-02, 7.6904e-03, -1.5747e-02, ..., 2.2888e-03, + 6.5002e-03, 5.9509e-03], + ..., + [ 2.8839e-03, 3.1128e-03, -2.5330e-03, ..., 2.6855e-03, + -7.0801e-03, 1.2390e-02], + [-1.1780e-02, -1.8539e-03, -1.4465e-02, ..., 5.3406e-03, + -9.0332e-03, 2.7466e-03], + [ 1.0010e-02, -1.4038e-03, 2.1820e-03, ..., -1.3855e-02, + 5.8594e-03, -1.8997e-03]], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_norm.weight': tensor([0.1064, 0.0157, 0.4277, 0.0698, 0.1680, 0.0364, 0.1484, 0.1348, 0.5938, + 0.0369, 0.0623, 0.0195, 0.0491, 0.0767, 0.0569, 0.1982, 0.4844, 0.0320, + 0.7422, 0.0349, 0.0435, 0.1680, 0.0178, 0.0796, 0.0481, 0.4043, 0.5508, + 0.2539, 0.0771, 0.0532, 0.0413, 0.2178, 0.6680, 0.2441, 0.0525, 0.0181, + 0.0102, 0.1099, 0.0159, 0.0466, 0.0488, 0.1299, 0.0183, 0.0403, 0.2041, + 0.0532, 0.3555, 0.0206, 0.0498, 0.0137, 0.0315, 0.3848, 0.5352, 0.3438, + 0.0903, 0.6484, 0.6016, 0.0398, 0.1235, 0.0325, 0.0854, 0.5586, 0.0106, + 0.5508, 0.1709, 0.3926, 0.0825, 0.0505, 0.0625, 0.4707, 0.0240, 0.5234, + 0.0212, 0.0356, 0.1338, 0.0869, 0.0286, 0.0898, 0.6367, 0.0347, 0.0879, + 0.0830, 0.0435, 0.0391, 0.4219, 0.0645, 0.0618, 0.0977, 0.0354, 0.1445, + 0.0469, 0.1748, 0.0537, 0.0596, 0.0298, 0.1216, 0.0131, 0.3613, 0.5820, + 0.0337, 0.0569, 0.2949, 0.0206, 0.1211, 0.0442, 0.2314, 0.0225, 0.6328, + 0.5859, 0.0325, 0.0119, 0.7695, 0.0187, 0.0140, 0.4961, 0.2070, 0.2754, + 0.7461, 0.0403, 0.0732, 0.0771, 0.5781, 0.6406, 0.0123, 0.0272, 0.2021, + 0.0261, 0.0210], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.v_proj.weight': tensor([[ 0.0135, -0.0027, -0.0303, ..., -0.0010, -0.0011, -0.0116], + [-0.0089, 0.0164, -0.0017, ..., 0.0066, -0.0045, -0.0234], + [ 0.0212, 0.0015, -0.0043, ..., 0.0004, -0.0052, -0.0410], + ..., + [-0.0072, 0.0032, 0.0109, ..., -0.0109, -0.0275, 0.0041], + [-0.0396, -0.0017, -0.0281, ..., -0.0083, 0.0090, -0.0190], + [-0.0095, -0.0029, 0.0029, ..., -0.0099, 0.0098, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.output_proj.weight': tensor([[ 0.0021, 0.0048, -0.0103, ..., -0.0038, 0.0046, 0.0111], + [-0.0361, 0.0028, -0.0017, ..., 0.0057, -0.0110, -0.0220], + [ 0.0036, -0.0078, -0.0037, ..., 0.0044, 0.0173, -0.0050], + ..., + [ 0.0079, -0.0135, 0.0210, ..., 0.0182, -0.0076, -0.0128], + [-0.0109, -0.0053, 0.0071, ..., 0.0124, -0.0144, 0.0092], + [-0.0041, -0.0201, 0.0154, ..., -0.0038, 0.0083, 0.0167]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0050, -0.0065, ..., 0.0266, -0.0278, 0.0078], + ..., + [-0.0217, -0.0410, 0.0488, ..., 0.0039, -0.0022, -0.0255], + [-0.0032, 0.0066, 0.0034, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0055, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, -0.0019, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0039, 0.0293, ..., 0.0065, 0.0089, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0140, 0.0126, -0.0093], + ..., + [-0.0044, 0.0320, -0.0233, ..., -0.0067, 0.0038, 0.0187], + [ 0.0136, -0.0352, 0.0267, ..., -0.0305, -0.0039, 0.0005], + [-0.0205, 0.0143, 0.0008, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0014, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0058, -0.0537, -0.0282, ..., -0.0109, -0.0146, -0.0023], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, 0.0029, 0.0015], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0024, ..., -0.0103, 0.0073, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.mlp.layer1.weight': tensor([[ 0.0123, 0.0036, 0.0047, ..., 0.0121, -0.0084, -0.0044], + [-0.0125, 0.0001, 0.0205, ..., 0.0079, -0.0267, -0.0065], + [ 0.0048, 0.0181, -0.0098, ..., 0.0228, -0.0131, 0.0217], + ..., + [-0.0101, 0.0042, 0.0571, ..., 0.0067, 0.0117, -0.0193], + [ 0.0087, -0.0047, -0.0078, ..., -0.0101, -0.0037, -0.0166], + [ 0.0046, 0.0154, -0.0159, ..., -0.0304, -0.0219, -0.0162]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.mlp.layer2.weight': tensor([[ 0.0048, 0.0209, -0.0045, ..., -0.0162, -0.0251, 0.0036], + [ 0.0038, 0.0232, 0.0094, ..., -0.0051, 0.0177, 0.0135], + [ 0.0050, -0.0222, 0.0045, ..., -0.0055, -0.0045, 0.0055], + ..., + [-0.0053, 0.0055, 0.0253, ..., -0.0049, -0.0100, 0.0018], + [ 0.0039, 0.0140, 0.0219, ..., -0.0047, 0.0172, 0.0111], + [-0.0066, 0.0059, -0.0039, ..., -0.0128, 0.0103, 0.0061]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0469, -0.0464, -0.0449, ..., 0.0010, 0.0008, 0.0009], + [-0.0386, -0.0381, -0.0371, ..., 0.0027, 0.0027, 0.0027], + [-0.0400, -0.0391, -0.0376, ..., 0.0022, 0.0023, 0.0023], + ..., + [ 0.0381, 0.0371, 0.0359, ..., -0.0036, -0.0037, -0.0036], + [ 0.0327, 0.0317, 0.0306, ..., -0.0031, -0.0030, -0.0030], + [-0.0386, -0.0376, -0.0371, ..., 0.0022, 0.0024, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0152, 0.0145, 0.0137, ..., 0.0220, 0.0248, 0.0148], + [ 0.0016, -0.0010, 0.0036, ..., 0.0025, 0.0036, 0.0003], + [-0.0003, -0.0022, -0.0009, ..., 0.0038, 0.0041, -0.0025], + ..., + [-0.0250, -0.0747, -0.0610, ..., 0.0364, 0.0369, -0.0679], + [ 0.0084, 0.0085, 0.0068, ..., 0.0199, 0.0176, 0.0087], + [ 0.0142, 0.0089, 0.0103, ..., 0.0192, 0.0216, 0.0101]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_cross_attn.1.weight': tensor([[-7.9102e-02, -7.6172e-02, -7.4707e-02, ..., -1.0252e-04, + -1.1730e-04, -1.0729e-04], + [ 3.0029e-02, 2.9297e-02, 2.8809e-02, ..., 9.4995e-07, + 2.6345e-05, 1.4687e-04], + [ 2.8442e-02, 2.7710e-02, 2.6489e-02, ..., 2.7466e-04, + 2.9755e-04, 2.0504e-04], + ..., + [ 6.4453e-02, 6.2988e-02, 6.2012e-02, ..., -1.8978e-04, + -1.8406e-04, -1.1444e-04], + [ 6.6895e-02, 6.4453e-02, 6.2988e-02, ..., -1.7643e-04, + -1.1778e-04, -1.6403e-04], + [-3.6133e-02, -3.5156e-02, -3.4424e-02, ..., -5.2929e-05, + 9.9659e-05, 1.2112e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0038, -0.0112, -0.0028, ..., 0.0201, 0.0288, 0.0041], + [ 0.0071, -0.0090, 0.0116, ..., 0.0005, 0.0060, -0.0019], + [ 0.0011, -0.0026, 0.0036, ..., -0.0082, -0.0055, -0.0093], + ..., + [-0.0141, 0.0030, 0.0330, ..., 0.0052, 0.0139, 0.0052], + [-0.0225, 0.0114, 0.0369, ..., 0.0081, 0.0171, -0.0190], + [ 0.0266, -0.0216, 0.0052, ..., -0.0014, 0.0077, 0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_mlp.1.weight': tensor([[-8.9844e-02, -8.6426e-02, -8.3984e-02, ..., 1.8787e-04, + 3.0899e-04, 2.5368e-04], + [ 2.2461e-02, 2.0264e-02, 1.9409e-02, ..., 1.8787e-04, + 2.6894e-04, 2.6321e-04], + [-9.2285e-02, -9.0332e-02, -8.9844e-02, ..., -4.0245e-04, + -3.2806e-04, -3.6621e-04], + ..., + [-8.2031e-02, -8.0566e-02, -7.9102e-02, ..., 2.1076e-04, + 3.1471e-04, 2.1172e-04], + [-6.1279e-02, -5.9570e-02, -5.8594e-02, ..., 2.2888e-04, + 9.2506e-05, 1.0252e-04], + [ 7.9102e-02, 7.7637e-02, 7.6660e-02, ..., -1.7738e-04, + -2.0599e-04, -1.6880e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_mlp.2.weight': tensor([[ 0.0023, 0.0038, 0.0038, ..., -0.0055, -0.0086, 0.0069], + [-0.0018, -0.0021, 0.0057, ..., -0.0016, 0.0037, 0.0015], + [-0.0010, 0.0045, -0.0019, ..., -0.0009, -0.0012, 0.0011], + ..., + [-0.0413, 0.0198, -0.0593, ..., -0.0295, -0.0106, 0.0225], + [ 0.0091, -0.0100, 0.0258, ..., 0.0067, 0.0037, -0.0063], + [ 0.0154, 0.0053, 0.0056, ..., 0.0135, 0.0070, -0.0107]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.after_proj.weight': tensor([[-0.0056, 0.0069, 0.0043, ..., 0.0067, 0.0135, -0.0127], + [-0.0050, -0.0069, 0.0068, ..., -0.0057, 0.0049, -0.0186], + [ 0.0104, -0.0056, 0.0124, ..., 0.0151, 0.0187, 0.0063], + ..., + [-0.0077, -0.0034, 0.0146, ..., 0.0031, -0.0150, -0.0096], + [-0.0122, -0.0050, -0.0278, ..., -0.0188, 0.0081, -0.0032], + [-0.0042, 0.0045, 0.0044, ..., -0.0145, 0.0129, 0.0003]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.after_proj.bias': tensor([ 0.0081, -0.0087, -0.0045, ..., -0.0044, -0.0139, -0.0038], + dtype=torch.bfloat16)} +[06-26 08:14:37|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:37|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 08:14:37|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 0 checkpoint. +[06-26 08:14:37|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/blur(ba2f44f2-c726-4fe7-949f-597069d9b91c) +[06-26 08:14:37|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 1 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/eb5325b77d358944da58a690157dd2b8071bbf85/general/blur/ba2f44f2-c726-4fe7-949f-597069d9b91c_ema_bf16.pt +[06-26 08:14:38|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0142, 0.0030, 0.0065], + [-0.0017, 0.0060, -0.0327, ..., -0.0070, 0.0009, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0135, -0.0033, 0.0283], + ..., + [-0.0031, 0.0125, 0.0248, ..., 0.0437, -0.0069, -0.0192], + [ 0.0277, 0.0146, 0.0040, ..., -0.0038, -0.0090, 0.0146], + [ 0.0437, 0.0366, 0.0085, ..., 0.0038, -0.0129, 0.0116]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0045, -0.0310, 0.0300, ..., -0.0068, -0.0048, -0.0339], + [-0.0103, -0.0048, 0.0115, ..., 0.0048, 0.0092, 0.0520], + [-0.0061, 0.0132, 0.0398, ..., -0.0096, -0.0179, -0.0061], + ..., + [-0.0046, 0.0175, -0.0095, ..., 0.0225, 0.0010, 0.0030], + [-0.0209, -0.0081, 0.0087, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0055, 0.0354, -0.0081, ..., 0.0118, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0011, ..., 0.0376, 0.0532, 0.0014], + [ 0.0175, -0.0347, -0.0013, ..., -0.0057, 0.0050, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0154, 0.0262, -0.0112], + [-0.0001, 0.0244, 0.0356, ..., 0.0019, 0.0170, 0.0119], + [ 0.0486, 0.0175, 0.0042, ..., -0.0020, -0.0023, -0.0021]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0141, 0.0569, 0.0115, ..., -0.0063, -0.0337, 0.0076], + [ 0.0491, 0.0413, 0.0251, ..., -0.0053, 0.0393, 0.0047], + [-0.0162, -0.0033, 0.0134, ..., -0.0198, -0.0244, 0.0469], + ..., + [ 0.0014, 0.0261, 0.0022, ..., -0.0041, 0.0347, -0.0033], + [-0.0254, 0.0068, -0.0016, ..., 0.0096, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0122, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0033, -0.0270, -0.0166, ..., 0.0118, -0.0298, 0.0532], + [-0.0031, -0.0035, -0.0128, ..., -0.0161, 0.0149, 0.0308], + [ 0.0144, 0.0086, 0.0146, ..., -0.0118, -0.0410, 0.0354], + ..., + [ 0.0062, -0.0630, -0.0034, ..., 0.0072, -0.0183, 0.0217], + [ 0.0038, 0.0036, 0.0258, ..., -0.0486, -0.0349, -0.0036], + [ 0.0182, -0.0081, 0.0197, ..., 0.0150, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0044, -0.0028, -0.0005, ..., 0.0003, 0.0015, -0.0123], + [-0.0150, 0.0251, 0.0432, ..., -0.0049, 0.0038, -0.0052], + [-0.0153, -0.0203, -0.0245, ..., -0.0120, -0.0325, 0.0040], + ..., + [ 0.0026, 0.0023, 0.0417, ..., 0.0280, 0.0030, -0.0042], + [-0.0024, 0.0001, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0065, -0.0019, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0155, -0.0062, -0.0481, ..., 0.0449, 0.0212, 0.0129], + [ 0.0154, -0.0227, 0.0220, ..., -0.0221, 0.0036, 0.0080], + [-0.0151, -0.0099, 0.0430, ..., -0.0095, 0.0327, -0.0111], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0139], + [-0.0017, 0.0166, -0.0059, ..., 0.0027, 0.0080, -0.0071], + [-0.0110, -0.0028, 0.0117, ..., 0.0058, -0.0034, -0.0075]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0040, -0.0386, -0.0118, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0079, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0126, -0.0044, -0.0126, ..., 0.0216, 0.0103, -0.0055], + ..., + [-0.0152, -0.0172, -0.0137, ..., 0.0098, 0.0098, 0.0184], + [-0.0152, -0.0356, 0.0201, ..., -0.0055, -0.0115, 0.0332], + [-0.0396, -0.0123, -0.0090, ..., -0.0284, -0.0142, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0277, 0.0312, 0.0133, ..., -0.0195, 0.0116, -0.0146], + [ 0.0192, 0.0095, -0.0075, ..., 0.0027, -0.0129, -0.0198], + [-0.0251, 0.0114, -0.0137, ..., 0.0093, 0.0172, 0.0075], + ..., + [ 0.0466, 0.0037, -0.0007, ..., -0.0044, 0.0239, 0.0031], + [ 0.0547, -0.0194, -0.0108, ..., -0.0056, 0.0311, -0.0023], + [ 0.0048, 0.0099, -0.0096, ..., 0.0203, 0.0170, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0045, -0.0032, 0.0283, ..., -0.0110, 0.0031, -0.0182], + [ 0.0049, 0.0129, -0.0315, ..., -0.0457, -0.0063, 0.0106], + [ 0.0027, 0.0261, 0.0245, ..., 0.0046, 0.0080, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0154, 0.0045, -0.0075], + [ 0.0055, -0.0371, 0.0356, ..., -0.0081, 0.0410, -0.0178], + [ 0.0178, -0.0151, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0079, 0.0208, 0.0273, ..., 0.0136, 0.0156, 0.0073], + [-0.0074, 0.0023, 0.0038, ..., 0.0146, 0.0015, 0.0050], + [ 0.0045, -0.0005, 0.0244, ..., -0.0120, -0.0055, -0.0247], + ..., + [ 0.0245, 0.0211, -0.0069, ..., 0.0247, -0.0092, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., -0.0052, -0.0295, 0.0400], + [-0.0116, 0.0132, 0.0305, ..., -0.0108, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0072, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0150, -0.0043, 0.0288, ..., 0.0141, 0.0583, -0.0025], + [-0.0281, 0.0156, -0.0386, ..., -0.0444, -0.0093, -0.0038], + ..., + [ 0.0041, 0.0157, 0.0217, ..., 0.0019, 0.0114, 0.0153], + [ 0.0059, 0.0195, -0.0283, ..., 0.0031, -0.0010, -0.0342], + [-0.0170, -0.0195, -0.0026, ..., 0.0049, 0.0165, 0.0133]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0139, -0.0146, 0.0459, ..., 0.0044, -0.0320, 0.0277], + [ 0.0033, 0.0064, 0.0270, ..., -0.0221, -0.0183, -0.0271], + [ 0.0221, 0.0219, -0.0201, ..., 0.0043, 0.0073, -0.0046], + ..., + [-0.0258, -0.0023, 0.0215, ..., 0.0041, 0.0237, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, 0.0014], + [ 0.0042, -0.0081, -0.0129, ..., -0.0334, 0.0060, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0145, -0.0454, ..., -0.0129, -0.0128, 0.0400], + [ 0.0295, -0.0187, -0.0109, ..., -0.0283, 0.0041, -0.0186], + [ 0.0052, 0.0547, 0.0033, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0280, -0.0058, -0.0269, ..., -0.0087, -0.0287, 0.0148], + [ 0.0091, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0098, 0.0104, ..., -0.0072, 0.0703, -0.0043]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0090, -0.0315, ..., 0.0078, 0.0254, 0.0359], + [ 0.0092, -0.0015, 0.0004, ..., -0.0123, -0.0233, -0.0046], + [-0.0025, 0.0046, -0.0126, ..., -0.0124, -0.0062, 0.0083], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0023], + [ 0.0339, -0.0072, 0.0223, ..., 0.0410, 0.0004, 0.0136], + [-0.0034, -0.0439, -0.0486, ..., 0.0139, -0.0129, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[-0.0035, 0.0141, -0.0068, ..., -0.0052, -0.0043, -0.0229], + [-0.0287, 0.0231, 0.0079, ..., -0.0142, -0.0036, 0.0222], + [ 0.0071, 0.0297, -0.0189, ..., -0.0168, 0.0311, -0.0262], + ..., + [ 0.0233, -0.0049, -0.0050, ..., 0.0045, 0.0278, -0.0034], + [-0.0103, 0.0035, 0.0142, ..., -0.0175, 0.0444, 0.0425], + [ 0.0078, -0.0079, 0.0099, ..., 0.0061, -0.0344, 0.0229]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[-0.0027, 0.0123, 0.0190, ..., -0.0096, -0.0198, 0.0148], + [-0.0581, 0.0294, 0.0032, ..., -0.0240, 0.0071, 0.0302], + [ 0.0087, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0040, -0.0289, 0.0537, ..., 0.0051, -0.0061, -0.0098], + [-0.0041, -0.0081, 0.0209, ..., 0.0049, 0.0101, 0.0186], + [-0.0087, -0.0349, -0.0173, ..., -0.0400, 0.0037, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0181, -0.0265, -0.0108, ..., 0.0096, -0.0133, -0.0405], + [ 0.0078, -0.0090, 0.0266, ..., -0.0400, 0.0073, 0.0391], + [-0.0045, 0.0359, 0.0284, ..., 0.0297, -0.0232, 0.0212], + ..., + [-0.0045, 0.0124, 0.0132, ..., -0.0134, 0.0016, 0.0095], + [ 0.0298, -0.0167, -0.0038, ..., -0.0070, 0.0048, -0.0206], + [ 0.0050, -0.0056, -0.0040, ..., 0.0009, -0.0044, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, 0.0002, 0.0123, ..., 0.0014, -0.0208, -0.0079], + [-0.0096, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0229], + [ 0.0137, -0.0278, 0.0117, ..., 0.0396, -0.0085, -0.0078], + ..., + [-0.0033, 0.0037, -0.0311, ..., 0.0344, 0.0062, -0.0124], + [-0.0236, -0.0100, -0.0067, ..., 0.0332, -0.0117, -0.0173], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0138, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[ 0.0038, 0.0166, 0.0041, ..., 0.0273, -0.0352, 0.0149], + [-0.0510, -0.0076, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0076, ..., -0.0388, 0.0004, 0.0398], + ..., + [-0.0160, -0.0195, -0.0052, ..., -0.0007, -0.0054, -0.0038], + [-0.0187, 0.0040, 0.0094, ..., 0.0287, -0.0122, -0.0098], + [-0.0181, -0.0038, 0.0247, ..., -0.0146, 0.0068, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0059, 0.0097, 0.0140, ..., -0.0112, -0.0019, -0.0253], + [-0.0166, 0.0330, -0.0171, ..., -0.0034, -0.0403, -0.0110], + [ 0.0084, 0.0347, 0.0275, ..., 0.0150, 0.0067, 0.0080], + ..., + [ 0.0271, -0.0167, -0.0106, ..., -0.0098, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0052, ..., 0.0033, 0.0017, -0.0243], + [-0.0079, 0.0581, 0.0067, ..., 0.0147, 0.0239, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0123, -0.0410, 0.0126, ..., -0.0254, -0.0027, -0.0173], + [ 0.0476, 0.0315, 0.0150, ..., -0.0146, -0.0070, 0.0038], + [-0.0092, -0.0270, 0.0086, ..., 0.0332, 0.0148, -0.0075], + ..., + [-0.0067, 0.0028, 0.0162, ..., -0.0177, 0.0171, -0.0134], + [ 0.0048, -0.0187, -0.0052, ..., 0.0248, -0.0140, -0.0058], + [-0.0142, 0.0022, 0.0150, ..., 0.0029, -0.0024, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0063, 0.0093, ..., 0.0255, -0.0253, -0.0132], + [-0.0048, -0.0135, -0.0160, ..., -0.0107, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0172, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0119, 0.0037, 0.0322, ..., -0.0183, -0.0033, 0.0031], + [ 0.0043, 0.0008, 0.0162, ..., 0.0028, 0.0255, 0.0071], + [ 0.0270, 0.0085, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0066, ..., -0.0532, 0.0104, -0.0042], + [-0.0240, -0.0215, -0.0103, ..., -0.0060, 0.0098, -0.0168], + [ 0.0062, -0.0046, -0.0064, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0059, 0.0172, -0.0106, ..., -0.0300, -0.0027, 0.0435], + [ 0.0049, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0066], + [-0.0167, 0.0123, 0.0101, ..., -0.0247, 0.0119, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0072, -0.0079, 0.0317, ..., 0.0199, -0.0105, -0.0376], + [ 0.0356, 0.0378, 0.0134, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., -0.0002, 0.0059, -0.0322], + ..., + [ 0.0059, 0.0107, -0.0040, ..., 0.0366, -0.0206, 0.0374], + [ 0.0024, -0.0079, 0.0040, ..., -0.0099, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0039, ..., -0.0166, -0.0059, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0048, -0.0087, -0.0344, ..., 0.0047, 0.0347, -0.0178], + [ 0.0267, -0.0366, -0.0089, ..., 0.0217, 0.0349, -0.0028], + [-0.0094, -0.0286, 0.0042, ..., -0.0026, -0.0282, 0.0029], + ..., + [-0.0128, 0.0037, 0.0053, ..., -0.0084, 0.0165, 0.0075], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0143], + [-0.0139, 0.0034, -0.0170, ..., -0.0128, 0.0061, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0028, 0.0034, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0118, 0.0374, ..., -0.0153, -0.0188, -0.0136], + ..., + [ 0.0262, -0.0019, -0.0075, ..., -0.0011, 0.0038, -0.0165], + [-0.0222, -0.0154, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0132, -0.0148, 0.0029, ..., 0.0020, -0.0234, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0102], + [ 0.0249, 0.0205, 0.0072, ..., -0.0032, -0.0092, -0.0103], + [-0.0220, -0.0022, -0.0023, ..., 0.0306, 0.0381, -0.0015], + ..., + [ 0.0200, -0.0011, 0.0170, ..., 0.0115, 0.0320, -0.0157], + [ 0.0349, 0.0117, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0067, 0.0256, -0.0023, ..., 0.0205, -0.0045, -0.0106]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0039, 0.0199, ..., 0.0013, 0.0276, -0.0312], + [ 0.0178, 0.0033, 0.0099, ..., 0.0110, -0.0398, 0.0293], + [ 0.0430, -0.0128, 0.0062, ..., 0.0040, -0.0084, 0.0067], + ..., + [ 0.0192, -0.0111, -0.0108, ..., -0.0045, 0.0330, 0.0140], + [-0.0126, -0.0542, -0.0205, ..., 0.0017, -0.0159, 0.0320], + [ 0.0303, 0.0066, -0.0425, ..., -0.0037, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0065, -0.0315, ..., 0.0525, 0.0039, -0.0349], + [-0.0019, -0.0039, -0.0154, ..., 0.0053, -0.0240, 0.0305], + [-0.0256, -0.0028, 0.0072, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0069, 0.0231, 0.0209, ..., 0.0075, 0.0045, 0.0461], + [-0.0072, -0.0079, -0.0001, ..., -0.0217, 0.0347, 0.0039], + [ 0.0032, 0.0234, 0.0038, ..., 0.0261, -0.0033, 0.0100]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0045, 0.0035], + [-0.0183, -0.0211, -0.0037, ..., -0.0145, 0.0138, -0.0162], + [-0.0066, -0.0160, 0.0070, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0096, 0.0078, ..., 0.0171, 0.0040, -0.0053], + [-0.0141, 0.0260, -0.0084, ..., 0.0078, 0.0103, -0.0166], + [ 0.0093, -0.0142, -0.0018, ..., 0.0053, 0.0056, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0070, 0.0141, -0.0053, ..., -0.0278, 0.0240, -0.0115], + [-0.0036, 0.0188, -0.0160, ..., -0.0317, -0.0043, 0.0181], + [-0.0006, 0.0310, -0.0085, ..., -0.0226, 0.0038, 0.0334], + ..., + [-0.0079, 0.0186, -0.0056, ..., 0.0228, 0.0052, 0.0046], + [ 0.0271, 0.0119, 0.0354, ..., 0.0019, 0.0198, 0.0143], + [ 0.0282, 0.0171, 0.0317, ..., -0.0110, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0114, -0.0265, -0.0110, ..., 0.0049, 0.0037, 0.0134], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0081], + [-0.0076, -0.0190, -0.0034, ..., 0.0166, -0.0135, 0.0137], + ..., + [-0.0200, 0.0018, 0.0060, ..., -0.0339, 0.0361, -0.0032], + [-0.0046, -0.0250, 0.0154, ..., -0.0022, -0.0162, 0.0220], + [-0.0044, 0.0038, 0.0205, ..., -0.0064, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0020, 0.0240, -0.0217, ..., 0.0172, -0.0028, 0.0115], + [-0.0386, 0.0103, 0.0074, ..., -0.0041, 0.0322, 0.0087], + [-0.0427, 0.0025, -0.0056, ..., 0.0068, 0.0110, -0.0139], + ..., + [-0.0084, -0.0043, -0.0264, ..., -0.0586, -0.0630, -0.0131], + [ 0.0064, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0106, -0.0610, 0.0162, ..., -0.0291, -0.0166, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0118, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0112], + [ 0.0108, -0.0101, -0.0327, ..., 0.0447, 0.0053, 0.0043], + [-0.0023, -0.0175, 0.0070, ..., -0.0076, -0.0049, 0.0026], + ..., + [ 0.0295, -0.0049, -0.0106, ..., 0.0095, -0.0160, -0.0288], + [ 0.0082, 0.0253, -0.0177, ..., 0.0095, 0.0071, 0.0066], + [-0.0135, -0.0057, 0.0137, ..., -0.0128, 0.0277, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0132, 0.0053, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0016, 0.0234, 0.0062, ..., 0.0464, 0.0069, 0.0309], + ..., + [-0.0118, -0.0175, 0.0327, ..., 0.0132, 0.0014, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, -0.0040, -0.0444, ..., 0.0060, -0.0283, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0081, -0.0101, 0.0146, ..., -0.0742, -0.0077, 0.0161], + [-0.0066, -0.0132, -0.0062, ..., 0.0039, -0.0023, 0.0238], + [ 0.0189, 0.0221, -0.0088, ..., 0.0018, 0.0408, 0.0045], + ..., + [ 0.0046, 0.0247, 0.0396, ..., 0.0043, -0.0304, -0.0183], + [ 0.0299, -0.0112, -0.0417, ..., -0.0039, -0.0376, -0.0045], + [-0.0210, 0.0168, -0.0050, ..., 0.0045, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0044, 0.0242, -0.0042, ..., -0.0245, 0.0033, -0.0070], + [ 0.0190, -0.0121, 0.0332, ..., -0.0144, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0037, ..., -0.0030, -0.0091, 0.0182], + ..., + [-0.0430, -0.0403, -0.0083, ..., 0.0132, -0.0141, 0.0055], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0090, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0050, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0247, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0093, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0061, ..., 0.0101, -0.0223, -0.0012], + ..., + [ 0.0192, 0.0292, -0.0018, ..., 0.0176, -0.0197, 0.0019], + [ 0.0265, 0.0063, 0.0056, ..., 0.0286, 0.0021, -0.0267], + [-0.0184, -0.0088, -0.0045, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0084, -0.0159, 0.0082, ..., 0.0198, 0.0051, 0.0277], + [ 0.0172, 0.0041, -0.0041, ..., -0.0003, 0.0040, 0.0067], + [ 0.0040, 0.0184, 0.0114, ..., 0.0099, -0.0053, -0.0059], + ..., + [ 0.0042, -0.0111, 0.0045, ..., -0.0096, 0.0036, 0.0121], + [-0.0427, -0.0391, -0.0192, ..., 0.0092, -0.0238, -0.0042], + [ 0.0038, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0083, -0.0161, 0.0123], + [ 0.0021, -0.0044, 0.0312, ..., 0.0071, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0071, 0.0046, 0.0278], + ..., + [ 0.0170, -0.0047, 0.0471, ..., -0.0057, 0.0038, -0.0079], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0039, -0.0203], + [ 0.0057, 0.0064, 0.0054, ..., -0.0095, 0.0013, 0.0021]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[ 0.0057, -0.0155, 0.0284, ..., -0.0104, 0.0135, -0.0052], + [ 0.0153, 0.0332, 0.0337, ..., 0.0223, -0.0165, 0.0038], + [-0.0020, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0069, 0.0062, 0.0280, ..., -0.0063, -0.0039, 0.0140], + [-0.0048, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0091, ..., -0.0035, 0.0167, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0045, ..., -0.0238, -0.0102, 0.0119], + [-0.0250, -0.0162, -0.0023, ..., -0.0084, 0.0002, -0.0369], + [ 0.0010, 0.0141, 0.0075, ..., -0.0452, -0.0045, 0.0120], + ..., + [ 0.0147, -0.0141, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0099, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0142, 0.0034, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[-0.0069, -0.0327, -0.0190, ..., 0.0042, 0.0036, -0.0209], + [ 0.0713, 0.0151, -0.0033, ..., 0.0051, 0.0186, 0.0020], + [-0.0227, -0.0092, 0.0188, ..., -0.0122, 0.0193, 0.0075], + ..., + [ 0.0116, 0.0464, 0.0083, ..., 0.0248, 0.0168, -0.0039], + [ 0.0381, -0.0036, 0.0060, ..., 0.0046, -0.0109, 0.0061], + [ 0.0292, 0.0050, 0.0209, ..., -0.0036, 0.0234, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0042, 0.0212, ..., 0.0305, 0.0137, 0.0104], + [-0.0060, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0065], + ..., + [ 0.0084, -0.0275, 0.0225, ..., -0.0039, -0.0189, -0.0107], + [-0.0125, 0.0041, -0.0045, ..., 0.0045, 0.0040, 0.0211], + [ 0.0020, -0.0060, -0.0059, ..., -0.0109, 0.0264, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[ 0.0043, -0.0076, -0.0108, ..., 0.0186, -0.0216, 0.0049], + [ 0.0039, 0.0123, -0.0231, ..., 0.0123, -0.0112, -0.0183], + [ 0.0107, 0.0103, -0.0011, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0123, 0.0146, ..., 0.0540, 0.0415, 0.0097], + [ 0.0038, 0.0087, -0.0138, ..., 0.0055, -0.0189, -0.0237], + [-0.0019, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, 0.0066, 0.0359, ..., -0.0084, -0.0121, -0.0041], + [ 0.0190, -0.0044, 0.0298, ..., 0.0126, 0.0048, -0.0200], + [ 0.0118, -0.0110, 0.0177, ..., -0.0248, 0.0229, -0.0038], + ..., + [-0.0077, -0.0048, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0025, -0.0291, ..., 0.0206, 0.0304, -0.0085], + [-0.0077, 0.0245, -0.0008, ..., -0.0413, 0.0216, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0060, -0.0157, -0.0032, ..., 0.0047, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0072], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, 0.0049], + ..., + [ 0.0033, -0.0058, 0.0258, ..., -0.0310, 0.0092, -0.0361], + [ 0.0040, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [-0.0036, 0.0157, -0.0046, ..., 0.0052, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0070, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0105, ..., 0.0104, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0037, ..., 0.0129, -0.0047, 0.0087], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [ 0.0038, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0128, 0.0206, ..., 0.0058, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0094, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0060, 0.0066, ..., 0.0015, 0.0112, 0.0598], + [-0.0060, 0.0267, -0.0065, ..., 0.0284, -0.0055, -0.0364], + ..., + [-0.0052, 0.0002, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, -0.0036, -0.0089, ..., 0.0211, 0.0134, -0.0078], + [ 0.0102, 0.0236, -0.0206, ..., -0.0134, 0.0048, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0077, 0.0294, ..., 0.0031, 0.0270, 0.0063], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0066, -0.0168], + [ 0.0201, -0.0121, 0.0126, ..., -0.0381, -0.0247, -0.0070], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0078], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0068, -0.0466], + [ 0.0327, 0.0284, -0.0039, ..., 0.0145, 0.0139, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, 0.0039, -0.0167, ..., 0.0005, 0.0170, 0.0107], + [ 0.0038, 0.0009, 0.0295, ..., -0.0172, 0.0038, 0.0135], + [ 0.0193, 0.0236, -0.0125, ..., -0.0045, -0.0369, 0.0239], + ..., + [ 0.0154, -0.0413, -0.0123, ..., -0.0354, 0.0062, -0.0320], + [-0.0266, -0.0071, -0.0327, ..., 0.0040, 0.0049, 0.0043], + [ 0.0298, 0.0356, -0.0192, ..., 0.0055, 0.0039, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0045, 0.0043, 0.0317, ..., 0.0053, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0071, -0.0096], + [-0.0166, 0.0059, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, 0.0049, 0.0260, ..., 0.0159, 0.0054, -0.0474], + [-0.0042, 0.0080, 0.0038, ..., 0.0143, 0.0557, 0.0023], + [-0.0203, 0.0273, 0.0029, ..., -0.0330, 0.0049, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0074, ..., 0.0036, -0.0093, -0.0277], + [ 0.0049, -0.0041, -0.0038, ..., -0.0049, 0.0140, -0.0161], + [-0.0049, 0.0168, -0.0057, ..., -0.0092, 0.0124, 0.0039], + ..., + [ 0.0398, 0.0452, 0.0251, ..., -0.0076, -0.0330, -0.0043], + [-0.0620, 0.0033, 0.0371, ..., 0.0410, 0.0037, -0.0173], + [ 0.0126, -0.0039, -0.0270, ..., -0.0037, 0.0388, -0.0104]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, -0.0039, 0.0245], + [-0.0449, 0.0039, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0061, ..., 0.0029, 0.0288, 0.0124], + [ 0.0330, 0.0040, 0.0059, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0074, -0.0030, ..., -0.0063, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0106, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0045, 0.0029, 0.0527], + [ 0.0041, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0086], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0095, -0.0095, 0.0327], + [ 0.0142, 0.0114, 0.0061, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0126, 0.0281, ..., -0.0242, -0.0126, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0029], + [-0.0053, 0.0039, -0.0286, ..., 0.0054, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0057, ..., -0.0037, 0.0081, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0096, 0.0150], + [-0.0025, 0.0021, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0071, ..., 0.0092, -0.0114, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0070, -0.0017, ..., -0.0220, -0.0145, 0.0040], + [-0.0166, 0.0189, -0.0065, ..., 0.0096, -0.0018, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0006, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0071, ..., -0.0041, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0070, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0122, 0.0289, 0.0125, ..., 0.0134, 0.0078, -0.0103], + [ 0.0188, -0.0197, -0.0039, ..., 0.0284, -0.0342, -0.0072], + [-0.0525, -0.0215, 0.0154, ..., -0.0270, 0.0084, 0.0254], + ..., + [-0.0068, 0.0087, 0.0281, ..., 0.0129, -0.0043, -0.0045], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0038, 0.0026], + [-0.0273, -0.0747, 0.0208, ..., 0.0041, 0.0041, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0047, 0.0150, 0.0334, ..., 0.0205, -0.0273, 0.0040], + [-0.0165, 0.0111, -0.0135, ..., 0.0106, -0.0069, 0.0219], + [ 0.0053, 0.0227, -0.0056, ..., 0.0128, -0.0070, 0.0150], + ..., + [-0.0347, -0.0148, 0.0061, ..., 0.0021, -0.0021, -0.0093], + [-0.0065, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0024, -0.0500, -0.0232, ..., 0.0320, 0.0022, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0059, -0.0364], + [ 0.0220, 0.0036, -0.0114, ..., 0.0047, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, -0.0039, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0074, 0.0192, -0.0393], + [-0.0101, -0.0244, 0.0275, ..., -0.0057, 0.0044, 0.0081], + [-0.0170, -0.0427, 0.0140, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0046, -0.0378, -0.0359, ..., 0.0552, -0.0073, -0.0160], + [ 0.0017, 0.0142, 0.0134, ..., 0.0049, -0.0153, 0.0354], + [ 0.0095, -0.0038, 0.0057, ..., -0.0016, -0.0249, -0.0135], + ..., + [ 0.0078, -0.0126, -0.0059, ..., -0.0095, -0.0084, -0.0028], + [-0.0059, 0.0439, 0.0166, ..., 0.0248, 0.0132, -0.0187], + [ 0.0175, 0.0060, 0.0043, ..., 0.0410, -0.0096, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0107, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0143, -0.0031, ..., -0.0039, -0.0244, 0.0356], + [ 0.0077, 0.0165, -0.0200, ..., -0.0048, 0.0237, -0.0071], + ..., + [-0.0197, -0.0022, 0.0386, ..., 0.0181, 0.0112, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, 0.0035, -0.0374], + [-0.0084, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0047, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0136, 0.0114, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, -0.0032, -0.0040, ..., -0.0048, -0.0030, 0.0337], + ..., + [ 0.0422, 0.0046, 0.0352, ..., -0.0304, -0.0039, -0.0204], + [ 0.0302, 0.0156, 0.0103, ..., -0.0076, -0.0254, -0.0245], + [-0.0041, -0.0069, 0.0042, ..., -0.0045, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, 0.0040, 0.0266, ..., 0.0154, 0.0187, 0.0039], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0041, ..., -0.0172, -0.0091, 0.0231], + ..., + [-0.0295, -0.0137, 0.0001, ..., -0.0041, -0.0052, 0.0099], + [-0.0045, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0075], + [ 0.0040, -0.0048, -0.0259, ..., 0.0045, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[-0.0047, 0.0062, -0.0140, ..., 0.0242, -0.0040, -0.0386], + [-0.0141, 0.0064, -0.0097, ..., -0.0369, -0.0155, 0.0157], + [-0.0267, -0.0020, -0.0068, ..., -0.0127, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0019, 0.0144, ..., 0.0115, 0.0039, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., 0.0041, -0.0039, -0.0137], + [-0.0219, -0.0396, -0.0078, ..., 0.0028, 0.0041, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0036, 0.0168, 0.0194, ..., 0.0349, -0.0420, 0.0013], + [ 0.0043, 0.0273, 0.0146, ..., -0.0184, -0.0035, 0.0154], + [ 0.0145, -0.0039, -0.0078, ..., 0.0245, 0.0359, 0.0059], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0037, 0.0044, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0073, -0.0312, -0.0457], + [ 0.0056, 0.0034, -0.0057, ..., 0.0021, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0037, 0.0121, 0.0039, ..., 0.0126, 0.0018, 0.0074], + [-0.0277, -0.0030, -0.0181, ..., 0.0038, -0.0029, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0043, 0.0513, -0.0113], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0113], + [-0.0282, -0.0042, 0.0121, ..., -0.0045, -0.0017, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0032, -0.0130, ..., 0.0141, 0.0308, 0.0056], + [ 0.0120, -0.0332, -0.0320, ..., -0.0073, -0.0142, -0.0032], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, 0.0045, -0.0311, ..., 0.0120, 0.0058, 0.0082], + [ 0.0043, 0.0198, 0.0141, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0125, -0.0297, -0.0055, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0103, ..., 0.0165, -0.0280, 0.0481], + [-0.0092, 0.0043, 0.0095, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, 0.0019, ..., 0.0302, -0.0025, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0039, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0022, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0048, -0.0222, 0.0262], + ..., + [-0.0147, 0.0049, 0.0260, ..., 0.0044, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0056, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0046, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0147, -0.0249, 0.0114], + [-0.0278, -0.0078, -0.0374, ..., 0.0276, -0.0062, 0.0150], + [ 0.0088, 0.0237, -0.0156, ..., 0.0527, -0.0075, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0072, -0.0095], + [-0.0071, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0034, -0.0129, ..., 0.0039, -0.0371, 0.0043], + [ 0.0059, -0.0127, 0.0043, ..., -0.0204, 0.0039, 0.0065], + [-0.0112, -0.0195, 0.0052, ..., -0.0542, -0.0089, -0.0045], + ..., + [-0.0205, -0.0084, 0.0038, ..., -0.0199, 0.0049, 0.0036], + [-0.0220, -0.0302, -0.0103, ..., 0.0107, 0.0369, -0.0038], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0039, 0.0049, 0.0238, ..., -0.0146, -0.0320, -0.0021], + [-0.0193, 0.0364, -0.0029, ..., -0.0061, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0063, -0.0232, -0.0464], + [-0.0110, -0.0298, -0.0195, ..., 0.0112, 0.0312, 0.0152], + [-0.0070, -0.0119, -0.0195, ..., -0.0042, 0.0109, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0060, -0.0239, -0.0123, ..., -0.0040, 0.0243, -0.0396], + [-0.0100, 0.0153, 0.0061, ..., -0.0277, -0.0045, 0.0016], + ..., + [ 0.0118, 0.0053, 0.0515, ..., 0.0327, -0.0118, -0.0064], + [-0.0040, 0.0170, -0.0006, ..., -0.0066, 0.0280, -0.0210], + [ 0.0012, 0.0075, 0.0041, ..., 0.0154, -0.0061, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0061, 0.0081, -0.0074, ..., -0.0588, 0.0187, -0.0067], + [-0.0103, 0.0199, -0.0078, ..., 0.0080, -0.0332, -0.0043], + ..., + [ 0.0045, -0.0464, 0.0277, ..., -0.0142, 0.0198, -0.0075], + [ 0.0162, 0.0075, 0.0089, ..., -0.0106, 0.0044, 0.0297], + [-0.0270, 0.0132, 0.0166, ..., 0.0081, 0.0175, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0053, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, -0.0007, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, -0.0039, -0.0201, ..., 0.0330, 0.0309, -0.0046], + ..., + [-0.0037, 0.0123, 0.0153, ..., -0.0042, -0.0037, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0039, 0.0175], + [-0.0244, -0.0058, -0.0300, ..., -0.0048, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0017, -0.0058, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, -0.0039, ..., 0.0310, 0.0088, -0.0037], + [ 0.0042, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0109, -0.0237, ..., -0.0104, 0.0157, 0.0064], + [ 0.0079, 0.0205, 0.0146, ..., 0.0262, -0.0195, 0.0082], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0080]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0029, ..., -0.0053, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., 0.0013, 0.0034, -0.0093], + [ 0.0017, 0.0002, 0.0398, ..., -0.0304, -0.0030, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0035, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0040], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0047, 0.0291], + [ 0.0132, -0.0035, -0.0339, ..., 0.0066, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, 0.0036], + [-0.0166, -0.0037, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0062, -0.0039, ..., 0.0240, -0.0037, -0.0031]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0007], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [ 0.0010, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0060, -0.0094], + [ 0.0059, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0078], + [-0.0206, 0.0417, -0.0125, ..., -0.0071, 0.0041, -0.0039], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0041, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, 0.0042, ..., -0.0164, 0.0036, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0028, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0054, -0.0231, -0.0255, ..., -0.0045, 0.0295, -0.0063], + [-0.0255, 0.0513, 0.0039, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0033, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0077], + [-0.0065, -0.0061, -0.0364, ..., -0.0175, 0.0306, -0.0077], + [-0.0051, -0.0195, 0.0019, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0075, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0067, -0.0193, -0.0140], + [-0.0166, -0.0048, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0128, -0.0076, 0.0140, ..., -0.0073, 0.0211, -0.0275], + [-0.0039, -0.0077, -0.0234, ..., 0.0087, -0.0070, -0.0166], + [ 0.0225, -0.0251, 0.0028, ..., 0.0078, 0.0236, 0.0081], + ..., + [ 0.0186, 0.0144, -0.0053, ..., 0.0272, -0.0084, -0.0092], + [-0.0040, -0.0030, 0.0071, ..., -0.0260, 0.0267, 0.0147], + [ 0.0101, 0.0009, 0.0212, ..., 0.0195, -0.0157, 0.0121]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0087, -0.0270, 0.0253, ..., 0.0099, 0.0048, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.1.proj.1.weight': tensor([[-0.0043, 0.0179, 0.0830, ..., -0.0645, 0.0269, -0.0361], + [-0.0288, -0.0048, -0.1138, ..., 0.0220, -0.0299, 0.0554], + [-0.0253, 0.1050, -0.0269, ..., -0.0189, 0.0175, -0.0332], + ..., + [ 0.0144, 0.0107, -0.0055, ..., 0.0226, -0.0234, 0.0046], + [-0.0284, -0.0649, 0.0503, ..., 0.0425, 0.0728, 0.0615], + [ 0.0162, -0.0171, 0.0017, ..., 0.0282, 0.0413, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_proj.weight': tensor([[-0.0054, -0.0018, 0.0101, ..., 0.0036, 0.0021, 0.0046], + [ 0.0051, 0.0034, -0.0007, ..., -0.0087, 0.0004, -0.0032], + [ 0.0102, -0.0041, 0.0022, ..., -0.0049, -0.0089, -0.0043], + ..., + [-0.0117, 0.0017, -0.0068, ..., 0.0032, 0.0063, -0.0047], + [ 0.0031, -0.0067, -0.0085, ..., -0.0039, -0.0055, -0.0053], + [ 0.0129, -0.0121, 0.0019, ..., 0.0051, -0.0067, 0.0019]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_norm.weight': tensor([ 1.5918e-01, -1.1572e-01, 1.1377e-01, 2.5977e-01, 7.2002e-05, + 5.5313e-04, 8.7172e-07, 7.4133e-07, 2.3041e-03, 4.2564e-10, + 1.2878e-09, -4.3165e-13, 1.2719e-12, 3.6948e-13, 4.2810e-13, + 7.7804e-13, 5.3478e-10, -1.0014e-05, -8.0327e-09, 3.6863e-17, + -1.7675e-13, 1.2612e-13, 2.4414e-02, 3.3379e-04, 2.8320e-02, + 6.4453e-01, 1.0400e-01, 4.0283e-03, 5.8838e-02, 9.5215e-02, + -1.3434e-14, -2.3761e-11, 6.0390e-10, -3.6380e-10, 8.9585e-11, + -6.4448e-07, 3.3833e-10, 2.2315e-14, 2.8848e-12, -1.6225e-09, + -5.2296e-12, 7.5033e-11, -5.2736e-15, 9.5367e-05, 2.4536e-02, + 6.1417e-04, -5.3955e-02, 5.8594e-02, 4.5117e-01, -1.0529e-03, + -1.8539e-03, 8.6799e-07, -3.5390e-08, -2.7008e-07, 2.2965e-11, + 9.8225e-10, -2.4556e-10, 1.8813e-07, 2.0227e-09, -2.2055e-11, + -2.7512e-18, -9.1073e-17, -3.8337e-16, -5.5422e-13, 2.7344e-01, + 2.4707e-01, 3.8086e-01, 5.0781e-02, -6.8545e-06, -3.0365e-03, + -6.7521e-09, 3.1710e-05, 2.0294e-03, -2.5437e-12, -1.0959e-10, + 3.0909e-13, 8.7766e-11, 1.1050e-10, 3.5749e-14, 1.7337e-12, + 1.9063e-09, 7.6834e-09, -3.8654e-11, -7.3830e-15, -3.1225e-15, + 1.7167e-11, 3.7537e-03, -1.1673e-03, 6.0156e-01, 7.9102e-02, + 5.3125e-01, 9.1553e-03, 9.8419e-04, 2.2888e-03, 1.9984e-14, + -1.2733e-09, -6.7812e-09, -9.9590e-11, -9.9681e-10, 5.9605e-06, + -3.2152e-13, 4.4409e-16, 2.5757e-14, 2.2847e-09, 2.1982e-14, + 7.9581e-11, -1.2506e-11, 1.7456e-02, -1.8234e-03, -1.6098e-03, + 4.7656e-01, 3.2227e-01, 7.0312e-02, -1.2665e-03, -1.2665e-03, + 1.6928e-05, -3.5288e-10, -7.1526e-07, -1.4097e-10, -6.1846e-10, + -5.4570e-11, 2.3865e-08, 3.2742e-11, -1.5250e-08, 5.2259e-17, + -1.6697e-17, 1.9451e-13, -1.4766e-14], dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.self_attn.k_proj.weight': tensor([[ 0.0060, 0.0034, 0.0004, ..., -0.0048, -0.0052, 0.0039], + [-0.0184, 0.0046, 0.0019, ..., 0.0067, -0.0027, 0.0020], + [ 0.0284, -0.0114, 0.0126, ..., -0.0076, 0.0044, 0.0008], + ..., + [-0.0018, 0.0026, -0.0046, ..., -0.0025, 0.0008, 0.0019], + [-0.0019, 0.0022, -0.0029, ..., -0.0026, 0.0020, 0.0021], + [-0.0021, 0.0010, -0.0025, ..., 0.0011, 0.0010, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.k_norm.weight': tensor([ 2.6758e-01, 1.9043e-01, 1.1426e-01, 2.2461e-01, 4.5538e-05, + -1.8234e-03, -8.0466e-07, -3.5018e-06, 2.4414e-03, 3.0195e-10, + 1.3461e-09, -5.6133e-13, -6.3949e-12, -1.7208e-14, 3.1619e-13, + -7.6739e-13, 6.0754e-10, 1.0133e-05, -6.8976e-09, -8.2399e-18, + 2.0517e-13, 1.1369e-13, 1.9165e-02, -6.6757e-04, 5.0293e-02, + 6.4453e-01, 1.2109e-01, -8.1177e-03, 3.9062e-03, 3.8605e-03, + -7.0083e-16, -2.1737e-10, -2.8667e-09, 3.5652e-10, 5.1841e-11, + 1.2368e-06, 3.3833e-10, -2.3093e-14, -2.9132e-12, 2.0227e-09, + -5.2864e-12, -9.0495e-11, 9.9365e-15, 1.2817e-02, 2.0264e-02, + -7.3433e-05, 1.2451e-01, 1.2061e-01, 4.7656e-01, -1.0376e-03, + 1.9684e-03, -6.7949e-06, 4.1677e-08, -5.0291e-07, 1.8616e-12, + 5.7116e-10, 1.3006e-10, -1.8626e-07, -2.0664e-09, -2.6375e-11, + -3.6863e-18, -2.9317e-16, -3.1398e-16, -5.5778e-13, 8.3496e-02, + 2.5391e-01, 3.8281e-01, -1.7383e-01, 3.1948e-05, -2.0905e-03, + -5.6578e-08, -2.1458e-06, -4.7302e-04, 1.9372e-10, 4.5929e-11, + -2.4514e-13, -7.7307e-11, 1.1914e-10, 6.5725e-14, 9.0949e-13, + -1.7026e-09, 1.0710e-08, -2.9786e-11, 2.7534e-14, 5.1903e-15, + 1.6485e-11, 2.5482e-03, -3.2043e-04, 5.9766e-01, -7.2266e-02, + 5.3125e-01, -4.3640e-03, 5.8105e-02, 9.4238e-02, -2.6201e-14, + 1.2005e-09, 5.5879e-09, 8.9130e-11, 1.0332e-09, -5.3644e-06, + -3.5882e-13, 1.4988e-15, -2.4691e-13, -2.3138e-09, -5.6399e-14, + -7.9581e-11, -1.2449e-11, 1.2390e-02, 1.6479e-02, 1.2360e-03, + 4.7656e-01, 2.9883e-01, 9.6680e-02, 9.9945e-04, -1.9264e-04, + 1.4544e-05, 3.4197e-09, 1.4668e-08, 1.3824e-10, -9.9681e-10, + -1.0823e-10, -2.7358e-08, -5.7128e-12, 1.5483e-08, 5.2692e-17, + -7.6978e-18, 1.9362e-13, -1.5721e-13], dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.self_attn.v_proj.weight': tensor([[ 0.0034, -0.0011, 0.0074, ..., -0.0056, 0.0005, -0.0013], + [ 0.0107, 0.0002, 0.0008, ..., 0.0023, -0.0015, -0.0035], + [ 0.0165, 0.0069, 0.0038, ..., -0.0015, 0.0008, 0.0028], + ..., + [ 0.0009, -0.0049, 0.0047, ..., 0.0081, -0.0025, -0.0019], + [-0.0013, -0.0104, -0.0045, ..., 0.0058, -0.0014, 0.0033], + [-0.0010, 0.0017, -0.0020, ..., -0.0033, 0.0019, -0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.output_proj.weight': tensor([[ 6.3171e-03, -2.3071e-02, -2.7222e-02, ..., -3.5858e-04, + -2.0599e-03, 8.5831e-04], + [-3.2043e-03, -8.5449e-03, 1.1353e-02, ..., 2.3499e-03, + 1.0925e-02, -5.2185e-03], + [ 4.8256e-04, 3.1128e-03, -2.5024e-03, ..., -8.8120e-04, + -5.2490e-03, 2.2888e-03], + ..., + [ 3.9673e-03, 5.1270e-03, 7.6294e-03, ..., 1.1444e-03, + -4.1809e-03, -7.2002e-05], + [-1.2085e-02, 7.8125e-03, 8.2397e-03, ..., -2.8687e-03, + 3.0212e-03, 1.2024e-02], + [-6.0425e-03, 4.6997e-03, -9.2773e-03, ..., -1.0071e-02, + 3.3528e-06, 5.1880e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_proj.weight': tensor([[-0.0041, -0.0047, -0.0041, ..., 0.0022, -0.0011, -0.0018], + [-0.0050, -0.0005, 0.0062, ..., 0.0010, -0.0060, -0.0040], + [ 0.0042, 0.0014, 0.0048, ..., -0.0018, 0.0037, 0.0048], + ..., + [-0.0006, 0.0045, -0.0013, ..., 0.0008, -0.0028, 0.0034], + [-0.0029, -0.0054, -0.0033, ..., -0.0016, -0.0026, -0.0128], + [ 0.0023, -0.0041, 0.0075, ..., -0.0037, 0.0009, 0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_norm.weight': tensor([ 1.7700e-03, 1.3447e-04, 1.0885e-08, 1.7319e-03, -2.4319e-05, + 2.3346e-03, -4.5896e-06, 4.3030e-03, 4.0283e-02, 2.7466e-02, + 3.2663e-05, 3.4424e-02, 5.6641e-01, 3.1586e-03, 2.4902e-02, + 1.1921e-04, 4.3750e-01, -1.3947e-05, 2.3041e-03, 5.0659e-03, + 1.1536e-02, 7.9727e-04, 9.2773e-02, 2.3041e-03, 1.0925e-02, + 4.1504e-03, 2.8687e-02, -1.3161e-04, 1.2268e-02, 1.9989e-03, + 3.6133e-01, 8.1787e-03, 2.6226e-05, 4.0820e-01, 5.3516e-01, + 3.3722e-03, 3.8910e-04, 3.8818e-02, 1.6895e-01, 3.1586e-03, + 3.3203e-02, 4.9133e-03, -4.6194e-07, 1.0010e-02, 4.0283e-03, + 3.8574e-02, 1.3794e-02, 1.0620e-02, 3.7689e-03, 9.6798e-05, + 9.5215e-03, 9.5963e-06, 1.9372e-06, 1.7262e-04, 2.2411e-04, + 2.5482e-03, 1.2207e-02, 8.6670e-03, 3.0518e-03, 8.3618e-03, + 5.8365e-04, -1.2779e-04, 1.0925e-02, 9.2163e-03, 2.0905e-03, + 1.2970e-03, 2.8992e-04, 2.6758e-01, 7.8125e-03, 1.6968e-02, + 3.9844e-01, 1.0586e-04, 4.3213e-02, 9.6130e-04, 3.6210e-06, + -1.0757e-03, 5.7068e-03, 8.3542e-04, 5.4932e-04, 1.3000e-02, + 3.7384e-03, 7.1106e-03, 1.8066e-01, 4.0894e-03, 1.0376e-03, + 2.4048e-02, 5.1270e-03, 4.1771e-04, 9.1171e-04, 4.7461e-01, + 4.7302e-03, 2.5330e-03, 1.4587e-02, -7.2956e-05, 2.7275e-04, + 4.2188e-01, 1.1396e-04, 2.2736e-03, -2.8558e-10, 1.1673e-03, + 4.0430e-01, 1.1683e-04, 9.8419e-04, 2.4223e-04, 7.5684e-03, + -1.0252e-04, 1.5106e-03, -2.0409e-04, 3.5248e-03, 5.7373e-03, + 1.1826e-03, 2.3365e-05, 8.4229e-03, 3.7231e-03, 5.0049e-03, + 5.3406e-04, 1.5564e-02, 1.8845e-03, 5.1270e-03, 5.2261e-04, + 7.3433e-05, 7.9956e-03, 5.4932e-03, 2.7847e-04, 1.3123e-03, + 3.4943e-03, 2.0409e-04, 3.8147e-04], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_proj.weight': tensor([[ 1.0452e-03, 7.2861e-04, 2.1057e-03, ..., -8.8120e-04, + -1.5717e-03, -8.1253e-04], + [-3.4790e-03, -2.0599e-03, -4.6997e-03, ..., -2.5787e-03, + -3.6621e-03, -2.4567e-03], + [-1.2512e-03, -1.4877e-03, -2.9907e-03, ..., -2.7466e-03, + -5.3406e-03, -2.0294e-03], + ..., + [ 2.0599e-03, -2.4719e-03, 2.5177e-03, ..., 1.4114e-03, + 1.0633e-04, 2.6703e-04], + [-8.0872e-04, 5.3024e-04, -2.4872e-03, ..., -3.3264e-03, + -5.9605e-05, 1.4877e-03], + [ 4.3106e-04, 1.8997e-03, -2.5024e-03, ..., -4.2114e-03, + 4.4632e-04, 1.0452e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_norm.weight': tensor([ 1.7624e-03, 1.3351e-04, -8.2946e-10, 1.4343e-03, 2.6703e-05, + 2.3499e-03, 5.7817e-06, 4.3030e-03, 3.9307e-02, 2.7466e-02, + 3.1710e-05, 3.4668e-02, 5.7031e-01, 3.1433e-03, 2.4414e-02, + 6.9141e-05, 4.3750e-01, -1.3828e-05, 2.1210e-03, 5.0049e-03, + 1.1536e-02, 8.9264e-04, 9.2773e-02, 2.4414e-03, 1.0864e-02, + 4.0894e-03, 2.8564e-02, -1.3065e-04, 1.1597e-02, 1.9989e-03, + 3.6133e-01, 7.7820e-03, 2.6464e-05, 4.0820e-01, 5.3516e-01, + 3.2654e-03, 5.8365e-04, 3.9062e-02, 1.6992e-01, 3.2959e-03, + 3.3203e-02, 4.8828e-03, -3.3900e-07, 9.8877e-03, 3.4943e-03, + 3.8574e-02, 1.3733e-02, 1.0559e-02, 3.7689e-03, 9.6321e-05, + 9.3994e-03, 9.8348e-06, 3.9339e-06, 1.8501e-04, 1.5869e-03, + 2.5787e-03, 1.2146e-02, 8.6670e-03, 3.5095e-03, 8.3618e-03, + 3.3379e-04, -1.2779e-04, 1.0925e-02, 9.0332e-03, 2.0905e-03, + 1.3351e-03, 2.8992e-04, 2.6758e-01, 7.7820e-03, 1.6968e-02, + 3.9844e-01, 1.0538e-04, 4.1260e-02, 8.8501e-04, 3.7402e-06, + 8.8501e-04, 4.8828e-03, 7.8964e-04, 5.4932e-04, 1.2878e-02, + 3.7689e-03, 7.0801e-03, 1.8066e-01, 3.8910e-03, 9.9945e-04, + 2.3926e-02, 6.3477e-03, 4.1771e-04, 8.8120e-04, 4.7461e-01, + 4.7607e-03, 2.4872e-03, 1.4282e-02, -7.2479e-05, 2.7275e-04, + 4.2188e-01, 6.0797e-05, 2.2583e-03, -3.5288e-10, 1.1597e-03, + 4.0430e-01, 1.2207e-04, 2.1362e-03, 2.3365e-04, 7.5378e-03, + -1.0347e-04, 1.5182e-03, -2.0409e-04, 2.8381e-03, 5.7373e-03, + 1.4877e-03, 3.7670e-05, 8.4229e-03, 3.6926e-03, 5.2185e-03, + -4.1199e-04, 1.5625e-02, 2.0142e-03, 6.0730e-03, 5.5313e-04, + -1.7548e-04, 8.0566e-03, 5.4321e-03, 1.6117e-04, 4.1425e-06, + 3.5095e-03, 2.9945e-04, 2.4319e-04], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.v_proj.weight': tensor([[-1.2390e-02, 3.1586e-03, 1.1902e-02, ..., 3.7842e-03, + 4.3640e-03, -1.0925e-02], + [-2.4658e-02, 2.7120e-06, 8.6060e-03, ..., 1.5030e-03, + -5.3101e-03, 6.1989e-05], + [-1.3000e-02, 3.0670e-03, 5.8594e-03, ..., 2.9602e-03, + 1.4648e-02, -1.2146e-02], + ..., + [ 9.6436e-03, 3.1090e-04, 1.8234e-03, ..., 1.4572e-03, + 5.8899e-03, 1.9043e-02], + [-9.1553e-03, 6.5613e-04, 1.0437e-02, ..., 1.2817e-03, + -1.7090e-02, -3.9368e-03], + [-1.5381e-02, -3.2654e-03, 5.3406e-03, ..., 1.8921e-03, + -8.4229e-03, -2.1362e-02]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.output_proj.weight': tensor([[ 0.0031, -0.0066, -0.0034, ..., -0.0017, 0.0056, -0.0074], + [ 0.0012, 0.0079, -0.0082, ..., 0.0134, 0.0061, -0.0087], + [-0.0018, -0.0019, -0.0030, ..., 0.0065, -0.0018, -0.0021], + ..., + [ 0.0019, 0.0023, 0.0046, ..., 0.0103, 0.0028, -0.0043], + [ 0.0082, -0.0067, 0.0049, ..., -0.0066, -0.0098, 0.0054], + [ 0.0154, -0.0197, -0.0110, ..., 0.0051, -0.0073, 0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_img.weight': tensor([[ 0.0057, 0.0117, -0.0008, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0077, 0.0249, 0.0291], + [-0.0092, 0.0061, -0.0121, ..., -0.0342, -0.0327, 0.0064], + ..., + [ 0.0070, -0.0040, 0.0016, ..., 0.0085, 0.0286, -0.0165], + [ 0.0071, -0.0060, -0.0100, ..., -0.0098, 0.0143, -0.0309], + [ 0.0242, -0.0104, 0.0255, ..., -0.0228, 0.0039, -0.0156]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0304, ..., -0.0220, 0.0247, 0.0033], + [ 0.0030, 0.0121, 0.0405, ..., 0.0128, -0.0410, 0.0361], + [-0.0089, 0.0113, -0.0087, ..., -0.0057, -0.0038, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0034, 0.0002], + [-0.0160, 0.0047, 0.0013, ..., 0.0134, -0.0175, -0.0044], + [-0.0026, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0102]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0048, -0.0364, ..., -0.0225, -0.0142, -0.0046], + [-0.0157, 0.0058, -0.0364, ..., 0.0099, 0.0119, 0.0079], + [-0.0040, 0.0615, 0.0044, ..., -0.0205, -0.0306, -0.0114], + ..., + [ 0.0038, -0.0040, -0.0103, ..., -0.0211, 0.0014, 0.0288], + [ 0.0266, -0.0117, -0.0052, ..., -0.0420, -0.0282, 0.0048], + [ 0.0286, 0.0325, 0.0097, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.mlp.layer1.weight': tensor([[ 5.4932e-03, -5.7373e-03, 4.5776e-03, ..., -4.4250e-03, + 3.2349e-03, -4.3945e-03], + [ 4.7607e-03, -3.5858e-03, -1.3351e-03, ..., 2.7008e-03, + -3.3722e-03, -3.7079e-03], + [ 3.9062e-03, 6.5308e-03, -1.2684e-04, ..., 9.3460e-04, + 2.3041e-03, -3.6469e-03], + ..., + [-5.8289e-03, -6.7444e-03, 3.8719e-04, ..., -9.0332e-03, + 6.6528e-03, -6.7749e-03], + [-2.0294e-03, 8.3618e-03, -5.5237e-03, ..., 6.8054e-03, + 1.2451e-02, -7.4463e-03], + [ 4.2419e-03, -9.6436e-03, -1.7700e-03, ..., 5.3883e-05, + -5.3101e-03, 1.6022e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.mlp.layer2.weight': tensor([[-0.0023, 0.0052, 0.0056, ..., 0.0024, 0.0047, -0.0038], + [-0.0092, 0.0043, 0.0011, ..., -0.0048, 0.0008, 0.0156], + [ 0.0038, 0.0040, -0.0080, ..., 0.0023, -0.0068, 0.0004], + ..., + [ 0.0035, -0.0021, 0.0041, ..., 0.0055, -0.0041, 0.0023], + [-0.0184, 0.0021, 0.0183, ..., -0.0038, 0.0152, -0.0055], + [ 0.0076, -0.0065, -0.0046, ..., 0.0020, 0.0037, -0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0496, -0.0488, -0.0474, ..., 0.0031, 0.0032, 0.0032], + [ 0.0510, 0.0500, 0.0486, ..., -0.0025, -0.0026, -0.0027], + [ 0.0503, 0.0496, 0.0486, ..., -0.0027, -0.0027, -0.0028], + ..., + [ 0.0483, 0.0476, 0.0459, ..., -0.0024, -0.0025, -0.0025], + [ 0.0449, 0.0439, 0.0425, ..., -0.0022, -0.0023, -0.0022], + [ 0.0498, 0.0486, 0.0476, ..., -0.0031, -0.0032, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0046, 0.0041, 0.0043, ..., 0.0056, 0.0057, 0.0046], + [-0.0015, 0.0024, 0.0016, ..., 0.0019, 0.0019, 0.0014], + [ 0.0006, -0.0019, -0.0016, ..., 0.0003, 0.0020, -0.0006], + ..., + [-0.0371, 0.0469, 0.0457, ..., 0.0430, 0.0391, 0.0378], + [-0.0066, 0.0071, 0.0073, ..., 0.0052, 0.0022, 0.0069], + [ 0.0089, -0.0101, -0.0099, ..., -0.0090, -0.0086, -0.0090]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_cross_attn.1.weight': tensor([[-6.6223e-03, -6.2561e-03, -5.4626e-03, ..., -6.3324e-04, + -5.6076e-04, -7.3242e-04], + [-3.2715e-02, -3.1250e-02, -3.1128e-02, ..., 2.0790e-04, + 2.1458e-04, 2.2507e-04], + [ 3.8330e-02, 3.7109e-02, 3.6621e-02, ..., -1.0395e-04, + -8.5831e-05, -1.0967e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1250e-02, ..., -2.8801e-04, + -2.2316e-04, -2.5940e-04], + [ 4.7607e-02, 4.6143e-02, 4.5166e-02, ..., -1.9455e-04, + -1.3447e-04, -1.9360e-04], + [ 3.4424e-02, 3.2959e-02, 3.2715e-02, ..., -1.8358e-05, + 5.8889e-05, 9.4771e-06]], dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_cross_attn.2.weight': tensor([[-1.4526e-02, -4.3335e-03, 3.7079e-03, ..., 4.5776e-03, + 4.6082e-03, 3.4637e-03], + [ 7.0190e-04, -4.8828e-04, 2.1973e-03, ..., 5.8746e-04, + 3.2654e-03, 2.2430e-03], + [ 4.6082e-03, -3.8605e-03, 3.8605e-03, ..., 3.8452e-03, + -7.1049e-05, 3.8605e-03], + ..., + [ 6.1340e-03, -5.5847e-03, 2.4292e-02, ..., 5.5847e-03, + 2.7100e-02, 1.9165e-02], + [ 9.7046e-03, 5.7983e-03, 1.1963e-02, ..., -5.6152e-03, + 1.3046e-03, 1.2085e-02], + [-1.3245e-02, -4.3640e-03, 5.9891e-04, ..., 4.1809e-03, + -1.2817e-03, 1.7262e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_mlp.1.weight': tensor([[-0.0072, -0.0069, -0.0063, ..., 0.0005, 0.0007, 0.0007], + [ 0.0537, 0.0520, 0.0505, ..., -0.0024, -0.0027, -0.0027], + [ 0.0425, 0.0415, 0.0405, ..., -0.0035, -0.0038, -0.0037], + ..., + [ 0.0874, 0.0850, 0.0835, ..., -0.0009, -0.0011, -0.0011], + [ 0.0054, 0.0049, 0.0030, ..., -0.0005, -0.0006, -0.0005], + [ 0.0530, 0.0518, 0.0503, ..., -0.0012, -0.0012, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_mlp.2.weight': tensor([[-0.0023, 0.0044, 0.0044, ..., 0.0054, 0.0023, 0.0067], + [-0.0008, 0.0017, 0.0017, ..., 0.0037, 0.0008, -0.0004], + [ 0.0019, 0.0015, 0.0017, ..., 0.0017, -0.0018, -0.0018], + ..., + [-0.0001, 0.0586, 0.0698, ..., 0.0488, 0.0009, 0.0188], + [ 0.0099, 0.0119, 0.0101, ..., 0.0149, -0.0098, 0.0068], + [-0.0082, 0.0038, 0.0038, ..., 0.0038, 0.0081, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.before_proj.weight': tensor([[ 5.5730e-06, -2.1362e-03, 2.8534e-03, ..., -4.4556e-03, + -2.5330e-03, -1.2894e-03], + [ 3.4485e-03, 9.2163e-03, -1.2589e-03, ..., 1.5335e-03, + 3.1586e-03, -3.7994e-03], + [-3.1586e-03, -3.9816e-05, 1.6098e-03, ..., -4.7302e-04, + 4.9210e-04, -1.7624e-03], + ..., + [-5.6839e-04, -1.8387e-03, -3.8757e-03, ..., 9.8419e-04, + 1.6251e-03, -1.9836e-03], + [-8.8215e-05, 6.7749e-03, 1.5411e-03, ..., 2.6131e-04, + 6.3782e-03, 2.7771e-03], + [-6.0120e-03, -5.4932e-03, 1.7700e-03, ..., 6.0654e-04, + 1.2817e-03, -2.1057e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.before_proj.bias': tensor([-0.0082, 0.0026, 0.0023, ..., -0.0035, -0.0018, 0.0038], + dtype=torch.bfloat16), 'net.control_blocks_1.0.after_proj.weight': tensor([[-0.0084, 0.0053, 0.0074, ..., -0.0014, 0.0119, 0.0034], + [-0.0170, 0.0038, -0.0052, ..., 0.0042, -0.0137, -0.0132], + [-0.0028, 0.0058, -0.0018, ..., -0.0056, 0.0006, -0.0009], + ..., + [-0.0042, 0.0040, -0.0036, ..., -0.0186, 0.0045, 0.0073], + [ 0.0078, -0.0033, -0.0013, ..., 0.0085, -0.0168, -0.0064], + [ 0.0062, -0.0062, -0.0072, ..., -0.0015, -0.0031, 0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.after_proj.bias': tensor([ 0.0045, -0.0116, -0.0005, ..., -0.0031, 0.0062, 0.0292], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_proj.weight': tensor([[-3.4424e-02, 9.1553e-03, -9.8267e-03, ..., 1.7334e-02, + 9.5749e-04, -9.5825e-03], + [ 1.3733e-02, -1.9165e-02, 4.2725e-03, ..., -2.0874e-02, + 9.9487e-03, -1.8799e-02], + [-5.3406e-03, -3.6163e-03, -5.1575e-03, ..., -5.1880e-03, + -3.4943e-03, 1.9531e-03], + ..., + [ 2.7924e-03, 3.7384e-03, 2.3193e-03, ..., 4.3030e-03, + 4.5471e-03, 6.9141e-05], + [-3.3264e-03, -3.7231e-03, -1.6403e-03, ..., -3.5706e-03, + 4.1389e-04, -1.7242e-03], + [ 5.6458e-03, -3.7842e-03, 3.2654e-03, ..., -1.7090e-03, + -2.4872e-03, 3.5248e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_norm.weight': tensor([ 2.4512e-01, 8.0859e-01, 1.5430e-01, 1.0469e+00, 8.6719e-01, + 1.5122e-34, 1.9861e-34, -1.9410e-34, 7.5984e-35, 5.8487e-07, + 1.0909e-34, 5.4543e-35, 2.5203e-35, -1.4294e-34, -1.6476e-34, + 9.7049e-35, 1.7676e-01, 2.2888e-04, 2.3079e-04, 9.5215e-03, + 3.4142e-04, 7.1411e-03, 4.1406e-01, 1.0703e+00, 1.1406e+00, + 4.1797e-01, 5.8838e-02, 5.2734e-01, -1.7822e-02, 1.6777e-34, + -4.0324e-34, 1.9034e-34, -3.2806e-03, 9.2697e-04, 4.7684e-04, + 3.9978e-03, 5.1498e-04, -1.5422e-34, -4.9277e-35, 1.3692e-34, + 1.3316e-34, -1.8161e-08, 1.3466e-34, 1.1094e+00, 5.5078e-01, + 4.0039e-01, 3.0664e-01, 1.2422e+00, 1.0547e-01, 1.2360e-03, + 1.1780e-02, 1.9043e-01, 2.6855e-03, -6.9965e-35, -1.6175e-34, + -1.8996e-35, 1.0193e-02, 2.7195e-07, 1.2488e-34, -8.4636e-35, + 3.8556e-35, 3.8744e-35, 1.0608e-34, -2.7836e-35, 5.5078e-01, + 2.7539e-01, 6.7578e-01, 3.6328e-01, 1.3574e-01, -9.5544e-35, + -6.8009e-34, 1.5046e-34, 4.5139e-35, 3.7998e-07, 1.1661e-34, + 1.9259e-34, 2.7836e-35, 7.5608e-35, -1.4219e-34, -1.3090e-34, + -9.3994e-03, 7.6675e-04, 4.4823e-04, 1.6632e-03, 9.3079e-04, + 4.8096e-02, 1.1016e+00, 4.3750e-01, 4.9023e-01, 1.1406e+00, + 2.8516e-01, 2.0117e-01, 6.5308e-03, -5.4769e-34, 2.6933e-34, + 2.0162e-34, 2.4986e-04, -2.4033e-04, 4.7874e-04, 4.5166e-03, + 8.5068e-04, 2.6519e-35, -4.1754e-35, 1.9259e-34, 1.1962e-34, + 1.2293e-07, -1.9259e-34, 3.0859e-01, 1.0625e+00, 1.1250e+00, + 1.2031e+00, 1.7773e-01, 7.6953e-01, 9.1553e-03, -1.9684e-03, + 2.1973e-03, 3.9978e-03, 2.8438e-34, -1.3015e-34, 3.3704e-34, + 9.9945e-04, 1.3828e-05, 6.9589e-35, 3.0394e-34, -1.1285e-35, + -2.1065e-34, 2.1535e-35, 6.3947e-35], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.self_attn.k_proj.weight': tensor([[ 0.0067, -0.0266, 0.0039, ..., -0.0156, -0.0106, -0.0172], + [ 0.0134, -0.0278, 0.0018, ..., -0.0325, 0.0129, -0.0195], + [ 0.0085, 0.0120, -0.0021, ..., -0.0072, -0.0024, -0.0022], + ..., + [ 0.0030, 0.0064, -0.0003, ..., 0.0005, 0.0013, -0.0032], + [-0.0025, -0.0030, -0.0014, ..., -0.0003, 0.0021, -0.0052], + [-0.0003, -0.0024, 0.0031, ..., -0.0048, 0.0014, -0.0027]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.k_norm.weight': tensor([ 4.0234e-01, 8.4766e-01, 2.1777e-01, 1.0625e+00, 8.6719e-01, + 5.9057e-35, 6.3947e-35, 6.7332e-35, -7.4855e-35, -2.3842e-07, + -1.0833e-34, -4.2882e-35, 4.4387e-35, -1.0006e-34, -1.4520e-34, + -1.3843e-34, 1.7676e-01, 3.7766e-04, 6.3896e-05, 4.6387e-02, + 3.3760e-04, -7.2937e-03, 5.1562e-01, 1.0703e+00, 1.1484e+00, + 3.4766e-01, 1.8066e-01, 6.1719e-01, -1.9653e-02, 2.7384e-34, + -4.8524e-35, 1.4896e-34, 3.4027e-03, 9.7656e-04, -1.9169e-04, + 4.6143e-02, 5.2643e-04, -5.0781e-35, -2.4262e-35, 8.5388e-35, + -7.7865e-35, -2.6659e-08, 1.6927e-34, 1.1094e+00, 4.1602e-01, + 1.2012e-01, -2.0605e-01, 1.2266e+00, 1.1377e-01, 1.6479e-03, + 9.2163e-03, 1.2354e-01, 2.1057e-03, 1.0457e-34, 2.0463e-34, + 4.3634e-35, 1.0132e-02, -3.3760e-08, -1.5723e-34, -7.8241e-35, + -1.2413e-35, -6.1314e-35, -1.4294e-34, -2.2569e-35, 4.9023e-01, + 3.3984e-01, 6.7969e-01, 2.5000e-01, 6.2500e-02, 1.5122e-34, + -1.0713e-33, 2.3623e-34, 2.3510e-35, -7.1526e-07, 2.4225e-34, + 2.3623e-34, 1.9184e-35, 8.0874e-35, 2.1535e-35, 1.0382e-34, + 1.7578e-02, 8.2779e-04, 1.6332e-05, 1.4267e-03, 9.3079e-04, + 4.7852e-02, 1.0938e+00, 3.0273e-01, 3.2617e-01, 1.1484e+00, + 2.3926e-01, 1.9434e-01, -2.1362e-04, -1.4520e-34, 2.5127e-34, + -2.9792e-34, -7.9155e-05, 4.1580e-04, 1.8978e-04, 4.5410e-02, + 8.5068e-04, 3.7428e-35, -4.5515e-35, -1.9259e-34, 8.5388e-35, + -1.2480e-07, -2.1516e-34, 3.5156e-01, 1.0781e+00, 1.1328e+00, + 1.2031e+00, -1.1182e-01, 7.8516e-01, 9.2163e-03, 6.6833e-03, + -1.6016e-01, -3.6011e-03, 4.3634e-34, -1.0608e-34, 4.9051e-34, + -1.6403e-03, 1.3828e-05, 2.9716e-35, 1.9184e-34, -3.7616e-35, + -1.0382e-34, -8.4259e-35, -4.7208e-35], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.self_attn.v_proj.weight': tensor([[ 0.0094, 0.0021, -0.0027, ..., -0.0101, -0.0006, 0.0021], + [ 0.0112, -0.0018, 0.0096, ..., 0.0132, -0.0034, -0.0018], + [ 0.0030, 0.0090, -0.0054, ..., 0.0119, 0.0067, -0.0090], + ..., + [ 0.0101, 0.0050, 0.0096, ..., 0.0025, -0.0022, -0.0150], + [ 0.0035, 0.0030, 0.0087, ..., 0.0302, -0.0042, 0.0026], + [-0.0038, -0.0031, -0.0037, ..., -0.0071, 0.0027, 0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.output_proj.weight': tensor([[-0.0118, -0.0044, -0.0154, ..., 0.0201, -0.0093, -0.0011], + [ 0.0154, 0.0047, 0.0190, ..., 0.0036, -0.0079, 0.0216], + [-0.0004, 0.0034, 0.0029, ..., -0.0008, 0.0037, 0.0030], + ..., + [-0.0063, 0.0133, 0.0054, ..., -0.0106, -0.0228, 0.0036], + [ 0.0079, -0.0027, -0.0053, ..., -0.0088, -0.0022, 0.0038], + [ 0.0028, 0.0053, -0.0009, ..., 0.0108, 0.0050, 0.0019]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_proj.weight': tensor([[-0.0126, 0.0129, 0.0119, ..., 0.0013, 0.0060, -0.0038], + [ 0.0101, -0.0017, 0.0041, ..., 0.0045, -0.0068, -0.0172], + [ 0.0040, 0.0110, -0.0029, ..., 0.0031, 0.0066, -0.0067], + ..., + [ 0.0084, -0.0129, 0.0032, ..., -0.0044, -0.0045, 0.0088], + [-0.0099, 0.0049, 0.0044, ..., 0.0073, 0.0020, 0.0093], + [ 0.0398, -0.0062, -0.0023, ..., 0.0098, 0.0096, -0.0177]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_norm.weight': tensor([ 2.2217e-02, 4.4922e-01, 2.2583e-02, 9.9487e-03, 4.8218e-03, + 8.4639e-06, 7.0572e-04, 5.1562e-01, 7.6675e-04, 3.9062e-01, + 2.9785e-02, 1.9653e-02, 8.4229e-03, 1.5015e-02, 1.8921e-02, + 1.4648e-02, 8.5449e-03, -9.6798e-05, 6.1512e-05, 3.1738e-02, + 2.6172e-01, 2.5635e-02, 5.4016e-03, 1.2512e-02, 6.2561e-03, + 1.2302e-04, -2.0266e-05, 4.8584e-02, 7.5102e-06, 3.7109e-01, + 2.0790e-04, 7.8082e-06, 9.6436e-03, -3.2336e-06, 1.0559e-02, + 7.2021e-03, 6.2466e-05, 3.9844e-01, 1.4343e-02, 6.5613e-03, + 7.0801e-02, 6.4392e-03, 1.0986e-02, 4.5471e-03, 1.2817e-02, + 8.6060e-03, 6.3171e-03, 1.7822e-02, 2.4902e-02, 1.3062e-02, + 1.4160e-02, 4.0234e-01, 1.1169e-02, 1.9897e-02, 3.1982e-02, + 1.3351e-03, 9.6680e-02, 1.3611e-02, 2.0264e-02, 1.4954e-02, + -5.2750e-06, 1.9043e-02, 1.5625e-02, 1.3053e-05, 1.2268e-02, + 8.2397e-03, 1.4496e-04, 8.3923e-04, 1.6357e-02, 7.8125e-02, + 1.2131e-03, 1.3123e-02, 4.8828e-02, 3.9795e-02, 1.6327e-03, + 1.8433e-02, -1.6212e-05, 2.5269e-02, 2.5635e-02, 2.5269e-02, + 6.8970e-03, 1.8555e-02, 3.4424e-02, -3.5286e-05, 2.1240e-02, + 3.3569e-03, 1.5030e-03, 4.4189e-02, 6.3782e-03, 1.4267e-03, + 1.6602e-02, 2.2030e-04, 7.4768e-03, 2.0020e-02, 9.0942e-03, + 5.9128e-05, 1.7456e-02, 4.4189e-02, 1.5991e-02, 4.2236e-02, + 3.8330e-02, 2.7588e-02, 1.2878e-02, 9.3460e-05, -3.5286e-05, + 5.5847e-03, 1.1414e-02, 2.1484e-02, 1.7456e-02, 2.5269e-02, + 7.3547e-03, 6.4087e-03, 3.5400e-02, 4.9023e-01, 3.0212e-03, + 1.6724e-02, 6.8054e-03, 1.1292e-02, 7.5989e-03, 1.0547e-01, + 1.4648e-03, 1.3733e-03, -9.6560e-06, 1.2634e-02, 5.9204e-03, + 2.1851e-02, 2.7466e-02, 1.6846e-02], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_proj.weight': tensor([[-0.0032, -0.0033, -0.0042, ..., -0.0037, -0.0011, 0.0017], + [ 0.0116, -0.0035, 0.0184, ..., 0.0208, 0.0198, 0.0116], + [-0.0008, 0.0036, 0.0039, ..., 0.0021, 0.0052, -0.0018], + ..., + [-0.0070, -0.0049, -0.0031, ..., -0.0012, 0.0021, -0.0016], + [-0.0072, -0.0053, -0.0142, ..., 0.0007, 0.0086, -0.0025], + [ 0.0007, -0.0004, -0.0055, ..., -0.0012, 0.0014, -0.0043]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_norm.weight': tensor([ 2.2827e-02, 4.4922e-01, 2.2461e-02, 9.9487e-03, 4.8523e-03, + -6.1095e-06, 7.2098e-04, 5.1562e-01, 8.0109e-04, 3.9062e-01, + 3.0518e-02, 1.9653e-02, 8.4229e-03, 1.5015e-02, 1.8921e-02, + 1.4526e-02, 9.0942e-03, -9.7275e-05, 5.2214e-05, 3.1982e-02, + 2.6172e-01, 2.4902e-02, 5.4321e-03, 1.2756e-02, 6.1951e-03, + 1.2493e-04, -1.8597e-05, 4.8096e-02, 6.1989e-06, 3.7109e-01, + 2.0885e-04, 9.1791e-06, 9.3384e-03, -1.7643e-05, 1.0559e-02, + 7.2632e-03, 6.2466e-05, 3.9844e-01, 1.4404e-02, 6.5002e-03, + 6.9824e-02, 6.4087e-03, 1.0803e-02, 4.5471e-03, 1.2756e-02, + 8.5449e-03, 6.4087e-03, 1.7700e-02, 2.4780e-02, 1.3306e-02, + 1.4160e-02, 4.0234e-01, 1.1169e-02, 1.9531e-02, 3.1982e-02, + 1.3275e-03, 9.6680e-02, 1.3611e-02, 1.9897e-02, 1.5076e-02, + -7.5102e-06, 1.9043e-02, 1.5625e-02, 1.3053e-05, 1.2207e-02, + 8.1177e-03, 1.4400e-04, 8.3542e-04, 1.6357e-02, 7.8125e-02, + 1.1597e-03, 1.3123e-02, 4.9072e-02, 3.9795e-02, 1.6632e-03, + 1.8555e-02, -2.3246e-05, 2.5024e-02, 2.5757e-02, 2.5269e-02, + 6.8970e-03, 1.8433e-02, 3.3936e-02, -3.6001e-05, 2.1118e-02, + 3.2501e-03, 1.4801e-03, 4.3945e-02, 6.6528e-03, 1.7242e-03, + 1.6968e-02, 2.5558e-04, 7.4158e-03, 1.9897e-02, 8.9111e-03, + 5.9128e-05, 1.6968e-02, 4.3701e-02, 1.5991e-02, 4.1992e-02, + 3.9062e-02, 2.7588e-02, 1.3000e-02, 9.0122e-05, -3.5524e-05, + 5.5542e-03, 1.1475e-02, 2.1240e-02, 1.7456e-02, 2.5146e-02, + 7.3547e-03, 6.1646e-03, 3.4912e-02, 4.8828e-01, 3.0670e-03, + 1.6724e-02, 6.8054e-03, 1.1169e-02, 7.7820e-03, 1.0547e-01, + 1.5488e-03, 1.3199e-03, -5.1856e-06, 1.3000e-02, 5.9814e-03, + 2.1973e-02, 2.7466e-02, 1.6724e-02], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.v_proj.weight': tensor([[-0.0178, -0.0001, -0.0044, ..., 0.0012, -0.0193, -0.0247], + [-0.0055, 0.0055, 0.0051, ..., 0.0037, 0.0095, 0.0070], + [-0.0400, -0.0034, -0.0039, ..., -0.0058, -0.0315, -0.0082], + ..., + [-0.0417, 0.0030, -0.0164, ..., 0.0028, 0.0205, 0.0055], + [ 0.0082, 0.0029, -0.0052, ..., 0.0016, 0.0115, 0.0044], + [ 0.0112, 0.0037, -0.0070, ..., 0.0033, 0.0131, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.output_proj.weight': tensor([[ 0.0064, 0.0042, 0.0162, ..., -0.0038, 0.0042, 0.0040], + [-0.0072, 0.0031, -0.0103, ..., -0.0123, 0.0031, 0.0117], + [-0.0011, 0.0020, -0.0045, ..., 0.0019, 0.0011, -0.0018], + ..., + [-0.0021, -0.0033, 0.0043, ..., -0.0018, 0.0036, 0.0046], + [ 0.0087, 0.0232, 0.0022, ..., -0.0349, 0.0075, 0.0289], + [ 0.0039, 0.0022, -0.0156, ..., -0.0105, 0.0062, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_img.weight': tensor([[ 0.0079, -0.0041, 0.0430, ..., -0.0327, 0.0048, 0.0144], + [-0.0278, 0.0280, 0.0249, ..., -0.0077, -0.0378, 0.0271], + [ 0.0139, -0.0119, 0.0312, ..., -0.0072, 0.0253, -0.0111], + ..., + [ 0.0245, -0.0152, 0.0043, ..., -0.0452, 0.0216, -0.0209], + [-0.0089, 0.0076, -0.0471, ..., -0.0107, 0.0157, -0.0083], + [ 0.0074, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0027, -0.0122, -0.0108], + [ 0.0135, 0.0048, -0.0188, ..., 0.0156, 0.0074, -0.0045], + [-0.0306, -0.0021, -0.0272, ..., 0.0103, 0.0337, 0.0017], + ..., + [-0.0034, -0.0042, 0.0049, ..., 0.0208, -0.0107, 0.0065], + [ 0.0193, 0.0210, 0.0122, ..., -0.0145, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0106, -0.0132, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img.weight': tensor([[ 0.0038, -0.0043, 0.0173, ..., 0.0083, 0.0356, 0.0217], + [-0.0498, -0.0033, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0108, 0.0043, -0.0139, ..., 0.0007, 0.0269, 0.0072], + ..., + [-0.0043, 0.0126, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0302, 0.0080, -0.0162, ..., -0.0050, -0.0074, -0.0118], + [ 0.0254, 0.0130, -0.0425, ..., -0.0264, 0.0138, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.mlp.layer1.weight': tensor([[ 0.0022, -0.0042, 0.0031, ..., 0.0037, 0.0016, -0.0054], + [-0.0031, 0.0053, 0.0071, ..., -0.0023, -0.0075, 0.0022], + [ 0.0026, 0.0027, 0.0036, ..., -0.0020, -0.0065, 0.0103], + ..., + [ 0.0281, 0.0128, -0.0095, ..., 0.0020, -0.0066, 0.0111], + [-0.0449, 0.0030, -0.0078, ..., 0.0183, 0.0075, -0.0037], + [ 0.0094, -0.0079, 0.0070, ..., -0.0075, 0.0014, -0.0247]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.mlp.layer2.weight': tensor([[-0.0096, -0.0112, 0.0197, ..., 0.0036, -0.0042, -0.0118], + [-0.0065, 0.0038, -0.0004, ..., -0.0036, -0.0001, -0.0032], + [ 0.0049, 0.0027, -0.0094, ..., -0.0126, 0.0036, 0.0025], + ..., + [ 0.0063, -0.0124, 0.0052, ..., -0.0036, -0.0082, 0.0097], + [ 0.0051, 0.0172, 0.0078, ..., 0.0063, 0.0067, 0.0023], + [ 0.0024, -0.0042, 0.0013, ..., -0.0013, -0.0005, 0.0079]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0410, -0.0398, -0.0393, ..., 0.0050, 0.0053, 0.0052], + [ 0.0327, 0.0315, 0.0305, ..., -0.0039, -0.0038, -0.0039], + [-0.0449, -0.0435, -0.0430, ..., 0.0051, 0.0052, 0.0052], + ..., + [-0.0454, -0.0442, -0.0430, ..., 0.0034, 0.0035, 0.0036], + [-0.0403, -0.0396, -0.0388, ..., 0.0037, 0.0039, 0.0039], + [-0.0231, -0.0222, -0.0211, ..., 0.0014, 0.0014, 0.0015]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_self_attn.2.weight': tensor([[ 6.8359e-03, -1.9302e-03, 5.8594e-03, ..., 6.8054e-03, + 6.8970e-03, 2.0142e-03], + [-1.6174e-03, 1.6403e-03, -1.7319e-03, ..., -2.3956e-03, + -1.7853e-03, -7.0572e-04], + [-1.0864e-02, -5.0735e-04, -9.8877e-03, ..., -1.1292e-02, + -1.0803e-02, 2.0862e-05], + ..., + [-1.8433e-02, -4.3678e-04, -2.4414e-02, ..., -2.6611e-02, + -2.5757e-02, 8.3542e-04], + [-1.0986e-02, -6.4697e-03, -1.1780e-02, ..., -1.1780e-02, + -1.0376e-02, 7.3242e-03], + [ 1.7395e-03, -6.0730e-03, 2.8534e-03, ..., 2.6550e-03, + 2.0447e-03, 5.0049e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5176e-02, 5.3711e-02, 5.3223e-02, ..., -1.0872e-04, + -1.5354e-04, -1.0729e-04], + [ 4.2114e-03, 4.1504e-03, 4.1504e-03, ..., -7.5340e-05, + -1.5163e-04, -2.6703e-04], + [-1.7700e-02, -1.6968e-02, -1.6235e-02, ..., -6.6757e-04, + -6.6757e-04, -6.7902e-04], + ..., + [-2.6001e-02, -2.6001e-02, -2.5024e-02, ..., -4.2343e-04, + -3.8147e-04, -3.4332e-04], + [-2.7100e-02, -2.6733e-02, -2.5635e-02, ..., -2.1839e-04, + -1.7262e-04, -2.2602e-04], + [ 5.4199e-02, 5.3223e-02, 5.2246e-02, ..., -1.7262e-04, + -5.4836e-05, -2.0218e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_cross_attn.2.weight': tensor([[-8.9111e-03, -4.9133e-03, -1.1353e-02, ..., -6.1646e-03, + 9.2163e-03, -9.5825e-03], + [ 9.6436e-03, -1.9150e-03, -3.1128e-03, ..., 2.9755e-03, + -8.4229e-03, 1.0071e-02], + [ 5.7983e-03, 1.8215e-04, -2.6941e-05, ..., 3.1281e-03, + -4.3335e-03, 5.3101e-03], + ..., + [ 6.0547e-02, 7.0190e-03, -4.3030e-03, ..., 5.7678e-03, + -1.6113e-02, 6.0059e-02], + [ 1.0620e-02, -4.6082e-03, -8.7891e-03, ..., 4.1504e-03, + -1.5625e-02, 1.1230e-02], + [-3.8605e-03, 9.1553e-03, -1.5503e-02, ..., 1.0376e-02, + 7.3242e-03, -3.8452e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_mlp.1.weight': tensor([[-2.8687e-02, -2.8564e-02, -2.8198e-02, ..., 3.9816e-05, + -4.6194e-06, 1.8477e-05], + [ 3.6621e-02, 3.6133e-02, 3.5889e-02, ..., -3.5248e-03, + -3.5553e-03, -3.4790e-03], + [ 3.9795e-02, 3.9062e-02, 3.8574e-02, ..., -3.6621e-03, + -3.6621e-03, -3.7079e-03], + ..., + [ 4.1016e-02, 3.9795e-02, 3.9551e-02, ..., -1.8616e-03, + -1.8845e-03, -1.8234e-03], + [ 1.9653e-02, 2.0264e-02, 2.0264e-02, ..., 1.6937e-03, + 1.6708e-03, 1.5640e-03], + [ 3.4912e-02, 3.4912e-02, 3.4668e-02, ..., -2.2583e-03, + -2.3193e-03, -2.3346e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_mlp.2.weight': tensor([[-0.0033, 0.0037, -0.0026, ..., 0.0036, 0.0035, 0.0038], + [ 0.0029, -0.0019, -0.0007, ..., -0.0013, -0.0033, -0.0018], + [ 0.0055, -0.0088, -0.0107, ..., -0.0078, -0.0045, -0.0078], + ..., + [ 0.0024, -0.0120, -0.0199, ..., -0.0101, 0.0016, -0.0090], + [ 0.0038, -0.0038, -0.0087, ..., -0.0038, -0.0038, -0.0038], + [ 0.0086, -0.0053, -0.0040, ..., -0.0063, -0.0093, -0.0060]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.after_proj.weight': tensor([[-3.2227e-02, 6.6528e-03, 3.7384e-03, ..., -1.4465e-02, + 7.3242e-03, -5.8899e-03], + [ 9.6436e-03, -6.9141e-05, 6.8054e-03, ..., 1.1902e-02, + 6.5918e-03, 1.8539e-03], + [ 1.4771e-02, -7.2021e-03, -1.2085e-02, ..., -4.2419e-03, + 5.5847e-03, 4.0894e-03], + ..., + [-7.9956e-03, 6.7139e-03, 4.1504e-03, ..., 1.5198e-02, + 6.6223e-03, 6.1340e-03], + [-1.2024e-02, -2.3499e-03, 5.7983e-03, ..., 5.2795e-03, + 6.1646e-03, 6.5613e-03], + [ 5.5847e-03, -7.5684e-03, 5.6458e-03, ..., 5.3406e-03, + -6.8359e-03, -1.2894e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.after_proj.bias': tensor([-0.0041, -0.0006, 0.0020, ..., -0.0030, -0.0014, -0.0011], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_proj.weight': tensor([[ 0.0452, -0.0189, -0.0077, ..., 0.0088, -0.0152, 0.0352], + [-0.0103, -0.0201, -0.0019, ..., -0.0052, 0.0085, -0.0058], + [ 0.0259, 0.0054, -0.0006, ..., 0.0060, -0.0057, 0.0159], + ..., + [-0.0143, -0.0048, 0.0012, ..., 0.0034, -0.0030, -0.0065], + [ 0.0223, 0.0277, 0.0019, ..., 0.0176, -0.0047, 0.0086], + [-0.0100, 0.0103, 0.0188, ..., -0.0136, -0.0041, -0.0189]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_norm.weight': tensor([ 1.4844e+00, 3.2617e-01, 4.1602e-01, 3.4570e-01, 1.3750e+00, + 2.7466e-04, 1.0437e-02, 5.0962e-06, 1.9775e-02, 5.3024e-04, + -9.7789e-08, -2.5705e-07, 1.1484e+00, 3.9978e-03, -3.8369e-13, + -1.6479e-03, -1.9379e-03, -2.3499e-03, 3.0396e-02, 1.5869e-02, + 1.5820e-01, -8.8692e-05, 4.2578e-01, 1.3594e+00, 5.0000e-01, + 5.5078e-01, 3.8477e-01, 1.6699e-01, 2.5513e-02, -2.5482e-03, + 1.3611e-02, -7.2098e-04, 1.4160e-02, 5.8899e-03, -1.2512e-03, + 3.4180e-02, 1.2988e-01, 5.9891e-04, -2.2430e-03, -4.0829e-06, + -1.2589e-04, 1.5182e-03, 2.2888e-03, 1.1562e+00, 9.0332e-02, + 1.3828e+00, 5.8203e-01, 2.0117e-01, -3.6865e-02, 4.9744e-03, + -2.3651e-03, 3.0273e-02, 1.3086e-01, 5.3048e-06, -7.1526e-05, + -1.1176e-06, -2.0981e-04, -6.3419e-05, -4.4678e-02, 4.8021e-09, + 3.5156e-02, -1.7262e-04, -2.9541e-02, -2.0020e-02, 6.0156e-01, + 1.4297e+00, 1.3281e+00, 1.3984e+00, 2.9688e-01, -1.6846e-02, + 1.3245e-02, 7.3016e-06, 2.1118e-02, -1.3367e-02, 1.4808e-07, + -2.3395e-06, 5.8105e-02, 2.7618e-03, 2.0073e-13, 8.1250e-01, + 1.0625e+00, 1.3000e-02, 1.9550e-04, 3.1982e-02, 1.1230e-01, + 9.8267e-03, 1.3359e+00, 3.1250e-01, 1.3672e+00, 1.4844e+00, + 2.2266e-01, 4.8047e-01, 4.7852e-02, -2.0630e-02, -5.2185e-03, + -6.7234e-05, -9.0332e-03, 1.4832e-02, -2.1515e-03, 3.3447e-02, + 1.2158e-01, 6.1989e-05, 1.1562e+00, 6.1095e-07, -4.6015e-05, + -3.7842e-03, 3.0100e-06, 7.1484e-01, 1.4219e+00, 1.3672e+00, + 1.5781e+00, 5.2734e-01, 2.7275e-04, 5.5542e-03, 6.3171e-03, + 3.1250e-02, 1.3770e-01, 1.4365e-05, -2.9504e-06, -6.2957e-07, + -6.6757e-05, -1.0452e-03, 1.1250e+00, 4.5262e-07, 2.8442e-02, + -1.5378e-05, 7.8125e-01, 1.1016e+00], dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.self_attn.k_proj.weight': tensor([[-0.0225, 0.0005, 0.0068, ..., 0.0205, 0.0119, 0.0415], + [-0.0079, 0.0108, -0.0090, ..., -0.0039, -0.0042, -0.0036], + [-0.0143, 0.0108, -0.0109, ..., 0.0058, 0.0182, 0.0007], + ..., + [-0.0030, -0.0069, 0.0044, ..., 0.0017, 0.0044, 0.0078], + [-0.0130, -0.0059, 0.0067, ..., 0.0051, 0.0056, 0.0075], + [ 0.0010, -0.0042, 0.0038, ..., -0.0051, -0.0101, 0.0044]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.k_norm.weight': tensor([ 1.5156e+00, 7.0703e-01, 3.8867e-01, 2.8711e-01, 1.3828e+00, + -7.5989e-03, 2.1484e-02, -4.3511e-06, 2.9663e-02, -1.7471e-03, + -1.0943e-07, -2.4959e-07, 1.1484e+00, 3.7537e-03, -3.7126e-13, + -1.6602e-02, -9.8419e-04, 8.0078e-02, 3.2227e-02, -3.7384e-03, + 5.3101e-03, 9.2030e-05, 5.8594e-01, 1.2344e+00, 8.6426e-02, + 5.4297e-01, 4.4141e-01, 1.6406e-01, 4.8828e-03, -5.6763e-03, + 1.4709e-02, -7.8201e-04, 1.3306e-02, 4.9561e-02, 1.9226e-03, + -1.7014e-03, 8.1787e-03, 5.9128e-04, 4.0894e-03, 4.0829e-06, + -1.1444e-04, 1.4114e-03, 2.2583e-03, 1.2891e+00, 6.6797e-01, + 1.3125e+00, 5.8203e-01, 2.9492e-01, -4.8828e-03, 6.2012e-02, + 3.7231e-03, -6.1951e-03, 3.3722e-03, -1.6928e-05, 6.5804e-05, + 1.1697e-06, -2.0885e-04, -3.1710e-05, -9.8877e-03, -3.7835e-09, + 5.4932e-03, -1.6975e-04, -3.5645e-02, -1.4099e-02, 7.0312e-01, + 1.3359e+00, 1.3359e+00, 1.3828e+00, 2.5977e-01, 1.5259e-02, + -2.2125e-03, 8.3447e-06, -2.8687e-03, -1.3184e-02, -1.3318e-07, + 2.3246e-06, 5.6763e-03, 3.0365e-03, 3.0376e-13, 8.1250e-01, + 1.0625e+00, 6.7871e-02, 1.3733e-04, -4.4861e-03, -4.0894e-03, + 9.9487e-03, 1.2500e+00, 7.3047e-01, 1.4766e+00, 1.5312e+00, + 4.5410e-02, 4.8828e-01, -1.4160e-02, 2.1118e-02, 1.6251e-03, + -6.7234e-05, 1.3794e-02, 5.3223e-02, 1.0071e-02, -5.9204e-03, + 8.5449e-03, -3.3617e-05, 1.1562e+00, 6.2212e-07, -8.0585e-05, + -3.8300e-03, 4.7386e-06, 3.7891e-01, 1.2891e+00, 1.3672e+00, + 1.6953e+00, 5.3125e-01, -2.5024e-02, 8.5938e-02, 6.5613e-03, + 1.7334e-02, -5.0964e-03, 1.1146e-05, 1.3709e-05, 5.3644e-07, + -2.5749e-05, -1.0529e-03, 1.1250e+00, -5.0664e-07, -7.2021e-03, + 2.2411e-05, 7.8125e-01, 1.1016e+00], dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.self_attn.v_proj.weight': tensor([[-0.0078, 0.0123, -0.0081, ..., -0.0261, 0.0009, 0.0117], + [ 0.0104, -0.0057, -0.0150, ..., 0.0039, -0.0039, 0.0118], + [-0.0068, -0.0043, 0.0041, ..., 0.0190, -0.0097, -0.0103], + ..., + [ 0.0154, -0.0128, -0.0153, ..., -0.0254, -0.0073, 0.0041], + [ 0.0048, -0.0039, 0.0100, ..., 0.0019, 0.0038, -0.0142], + [-0.0115, 0.0057, -0.0239, ..., 0.0041, 0.0121, 0.0092]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.output_proj.weight': tensor([[ 0.0040, 0.0135, -0.0322, ..., 0.0049, -0.0205, 0.0128], + [ 0.0024, -0.0211, -0.0120, ..., 0.0063, 0.0154, -0.0089], + [-0.0041, 0.0023, 0.0042, ..., 0.0028, 0.0031, 0.0042], + ..., + [ 0.0118, 0.0081, -0.0229, ..., 0.0190, 0.0114, 0.0009], + [ 0.0064, 0.0043, 0.0041, ..., 0.0077, 0.0093, 0.0035], + [ 0.0062, 0.0047, -0.0041, ..., -0.0068, 0.0107, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_proj.weight': tensor([[-0.0069, 0.0082, 0.0011, ..., -0.0103, -0.0052, 0.0002], + [ 0.0091, 0.0072, 0.0038, ..., -0.0041, -0.0059, 0.0029], + [ 0.0051, 0.0062, -0.0024, ..., -0.0035, 0.0157, -0.0038], + ..., + [ 0.0027, 0.0070, 0.0029, ..., 0.0079, 0.0060, 0.0046], + [-0.0107, 0.0010, 0.0054, ..., 0.0030, -0.0099, 0.0085], + [ 0.0070, 0.0062, -0.0022, ..., -0.0051, 0.0134, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_norm.weight': tensor([3.5400e-03, 1.4551e-01, 1.3245e-02, 5.8594e-01, 2.1240e-02, 2.3315e-02, + 8.7891e-03, 4.8523e-03, 1.7456e-02, 1.2598e-01, 8.4839e-03, 5.5664e-02, + 2.7344e-01, 1.9165e-02, 1.8945e-01, 1.5320e-02, 6.7871e-02, 1.5442e-02, + 4.1504e-03, 1.8555e-02, 1.0559e-02, 1.8652e-01, 2.8711e-01, 1.9922e-01, + 2.3315e-02, 1.5918e-01, 6.6406e-02, 1.6357e-02, 4.0625e-01, 8.6670e-03, + 1.9775e-02, 5.1172e-01, 6.6895e-02, 3.4668e-02, 1.5991e-02, 1.5991e-02, + 7.5195e-02, 4.3359e-01, 5.3125e-01, 1.2451e-02, 8.6060e-03, 1.2109e-01, + 5.6152e-03, 5.2344e-01, 3.0273e-02, 5.0293e-02, 1.6235e-02, 3.8086e-01, + 7.6294e-03, 1.8555e-02, 8.2520e-02, 1.2451e-02, 1.8799e-02, 4.6484e-01, + 1.8768e-03, 2.6978e-02, 5.4932e-02, 8.0078e-02, 6.2256e-03, 1.5991e-02, + 4.8523e-03, 2.3926e-02, 1.0559e-02, 3.6133e-01, 1.6602e-02, 3.0670e-03, + 2.1118e-02, 5.9570e-02, 5.9814e-02, 1.3611e-02, 6.6528e-03, 6.0156e-01, + 3.7575e-04, 5.7031e-01, 1.3855e-02, 3.9673e-03, 1.6602e-02, 3.0664e-01, + 2.6733e-02, 3.3722e-03, 1.0986e-01, 5.2490e-03, 6.3477e-02, 9.1553e-03, + 1.9897e-02, 8.1787e-03, 1.2390e-02, 6.5430e-02, 2.0508e-01, 1.0315e-02, + 1.5564e-03, 1.8433e-02, 4.6387e-02, 2.4170e-02, 2.6245e-02, 1.5747e-02, + 1.7212e-02, 2.3193e-02, 6.3965e-02, 2.7344e-02, 4.3213e-02, 1.5198e-02, + 3.1738e-03, 6.5430e-02, 5.8594e-01, 7.6172e-02, 6.8359e-03, 4.8047e-01, + 1.3428e-02, 3.0762e-02, 1.7456e-02, 1.3000e-02, 1.4591e-04, 9.0332e-02, + 4.7852e-01, 2.4170e-02, 2.4292e-02, 5.7422e-01, 1.2158e-01, 2.6001e-02, + 1.9775e-02, 6.8750e-01, 1.5234e-01, 1.5869e-02, 1.4404e-02, 5.4932e-03, + 2.8442e-02, 1.1292e-02], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_proj.weight': tensor([[ 2.1057e-03, -3.3875e-03, 4.6692e-03, ..., 6.9427e-04, + -4.0283e-03, 4.7302e-03], + [-7.0190e-03, 2.6321e-04, -5.3711e-03, ..., 5.0964e-03, + -7.8125e-03, -5.4626e-03], + [-5.3711e-03, -1.6174e-03, -5.2490e-03, ..., 1.2451e-02, + -1.4526e-02, -3.6926e-03], + ..., + [ 6.9580e-03, 4.6082e-03, 2.5482e-03, ..., 2.9297e-03, + -1.8768e-03, -6.6223e-03], + [ 6.2866e-03, 2.9449e-03, 1.3504e-03, ..., -1.3275e-03, + 2.6398e-03, 4.4250e-03], + [ 6.9141e-05, 6.7902e-04, -1.2589e-03, ..., -6.3782e-03, + -6.1340e-03, 3.0823e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_norm.weight': tensor([ 3.5400e-03, 1.4551e-01, 1.3062e-02, 5.8594e-01, 2.0996e-02, + 2.3193e-02, 8.9722e-03, 4.7302e-03, 1.7700e-02, 1.2598e-01, + 8.4839e-03, 5.3955e-02, 2.7344e-01, 1.9287e-02, 1.8945e-01, + 1.6113e-02, 6.7383e-02, 1.5442e-02, 4.0894e-03, 1.8555e-02, + 1.0803e-02, 1.8750e-01, 2.8906e-01, 1.9922e-01, 2.3071e-02, + 1.5918e-01, 6.5430e-02, 1.5137e-02, 4.0820e-01, 8.6670e-03, + 2.0020e-02, 5.1172e-01, 6.6895e-02, 3.4424e-02, 1.6357e-02, + 1.5625e-02, 7.5195e-02, 4.3359e-01, 5.3125e-01, 1.2451e-02, + 8.5449e-03, 1.2109e-01, 5.5847e-03, 5.2344e-01, 3.0762e-02, + 5.0293e-02, 1.6479e-02, 3.8086e-01, 7.7209e-03, 1.8433e-02, + 8.2520e-02, 1.2817e-02, 1.8799e-02, 4.6289e-01, 2.1057e-03, + 2.6978e-02, 5.4688e-02, 8.0078e-02, 6.3171e-03, 1.4343e-02, + 4.7913e-03, 2.3804e-02, 1.0681e-02, 3.6133e-01, 1.6479e-02, + 3.1281e-03, 2.0630e-02, 5.9570e-02, 5.9814e-02, 1.3611e-02, + 7.0496e-03, 6.0156e-01, 1.6022e-04, 5.7031e-01, 1.4038e-02, + -4.2114e-03, 1.6602e-02, 3.0664e-01, 2.7344e-02, 3.4637e-03, + 1.1035e-01, 5.0964e-03, 6.2988e-02, 8.8501e-03, 1.9409e-02, + 8.2397e-03, 1.3000e-02, 6.5918e-02, 2.0508e-01, 1.0193e-02, + 1.7548e-03, 1.7212e-02, 4.6631e-02, 2.4170e-02, 2.6978e-02, + 1.5869e-02, 1.7334e-02, 2.3193e-02, 6.3965e-02, 2.7344e-02, + 4.2725e-02, 1.5320e-02, 3.8452e-03, 6.5918e-02, 5.8594e-01, + 7.6172e-02, 6.6223e-03, 4.8047e-01, 1.3062e-02, 3.0518e-02, + 1.8066e-02, 1.2939e-02, 1.5259e-04, 9.0332e-02, 4.7852e-01, + 2.3682e-02, 2.4292e-02, 5.7422e-01, 1.2305e-01, 2.6245e-02, + 1.9653e-02, 6.9141e-01, 1.5137e-01, 1.4648e-02, 1.3916e-02, + 5.4321e-03, 2.7954e-02, 1.1108e-02], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.v_proj.weight': tensor([[ 0.0136, -0.0051, -0.0161, ..., 0.0037, -0.0212, -0.0067], + [-0.0162, -0.0009, -0.0035, ..., -0.0041, -0.0212, 0.0166], + [-0.0165, -0.0024, 0.0033, ..., 0.0016, 0.0272, 0.0325], + ..., + [ 0.0039, -0.0062, -0.0161, ..., 0.0018, 0.0115, 0.0039], + [ 0.0231, 0.0065, 0.0222, ..., 0.0047, 0.0070, 0.0078], + [ 0.0183, 0.0043, 0.0031, ..., 0.0067, -0.0141, -0.0238]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.output_proj.weight': tensor([[ 0.0136, 0.0078, -0.0025, ..., -0.0070, 0.0131, 0.0087], + [-0.0044, -0.0047, 0.0051, ..., 0.0121, 0.0042, -0.0007], + [ 0.0028, -0.0032, 0.0048, ..., -0.0025, 0.0024, -0.0036], + ..., + [-0.0141, -0.0090, -0.0069, ..., 0.0048, -0.0078, 0.0093], + [ 0.0165, -0.0115, 0.0109, ..., -0.0265, -0.0118, -0.0236], + [ 0.0045, -0.0073, 0.0156, ..., 0.0239, 0.0071, 0.0122]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0066, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [ 0.0018, 0.0150, -0.0159, ..., 0.0040, -0.0058, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0070, -0.0072], + [-0.0337, -0.0037, 0.0052, ..., -0.0039, 0.0172, 0.0231], + [-0.0255, -0.0015, -0.0039, ..., 0.0408, 0.0430, -0.0123]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0031], + [ 0.0198, -0.0208, 0.0129, ..., 0.0088, -0.0337, -0.0051], + [ 0.0302, 0.0101, 0.0025, ..., -0.0304, 0.0199, -0.0019], + ..., + [ 0.0171, -0.0276, -0.0075, ..., 0.0231, 0.0182, -0.0045], + [ 0.0221, 0.0388, 0.0060, ..., 0.0037, -0.0488, -0.0361], + [ 0.0040, 0.0115, -0.0058, ..., 0.0037, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img.weight': tensor([[ 0.0273, 0.0038, 0.0118, ..., -0.0015, 0.0347, 0.0615], + [-0.0258, 0.0009, -0.0197, ..., 0.0139, -0.0045, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0096, 0.0132, 0.0184, ..., 0.0110, 0.0084, 0.0515], + [-0.0010, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.mlp.layer1.weight': tensor([[-0.0015, 0.0028, 0.0081, ..., 0.0044, -0.0105, -0.0049], + [-0.0078, 0.0064, 0.0078, ..., -0.0059, -0.0128, 0.0036], + [-0.0074, 0.0156, 0.0066, ..., 0.0040, -0.0019, -0.0508], + ..., + [-0.0068, 0.0077, -0.0031, ..., 0.0070, 0.0021, -0.0077], + [ 0.0013, -0.0021, -0.0081, ..., -0.0139, -0.0040, -0.0269], + [ 0.0014, 0.0190, 0.0085, ..., 0.0118, 0.0143, -0.0080]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.mlp.layer2.weight': tensor([[-0.0039, -0.0052, -0.0156, ..., -0.0110, -0.0114, 0.0038], + [ 0.0078, -0.0022, 0.0026, ..., -0.0114, -0.0106, 0.0042], + [-0.0045, 0.0028, -0.0003, ..., 0.0067, 0.0045, 0.0061], + ..., + [-0.0017, -0.0100, -0.0094, ..., 0.0044, 0.0038, 0.0044], + [-0.0137, -0.0057, 0.0025, ..., 0.0140, 0.0033, -0.0125], + [ 0.0079, 0.0063, -0.0026, ..., -0.0060, 0.0133, -0.0347]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0299, 0.0299, 0.0286, ..., -0.0041, -0.0042, -0.0043], + [-0.0284, -0.0275, -0.0271, ..., 0.0042, 0.0046, 0.0045], + [-0.0298, -0.0292, -0.0287, ..., 0.0044, 0.0048, 0.0048], + ..., + [-0.0344, -0.0332, -0.0327, ..., 0.0020, 0.0021, 0.0023], + [-0.0337, -0.0332, -0.0325, ..., 0.0020, 0.0021, 0.0020], + [-0.0156, -0.0157, -0.0159, ..., 0.0001, 0.0001, 0.0001]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0055, -0.0037, -0.0053, ..., -0.0017, -0.0019, -0.0014], + [ 0.0044, -0.0040, -0.0042, ..., -0.0033, -0.0038, -0.0022], + [ 0.0057, -0.0066, -0.0056, ..., -0.0018, -0.0021, 0.0015], + ..., + [ 0.0420, -0.0452, -0.0454, ..., -0.0278, -0.0244, 0.0075], + [-0.0060, 0.0142, 0.0075, ..., 0.0167, 0.0153, 0.0170], + [-0.0189, 0.0197, 0.0195, ..., 0.0182, 0.0177, 0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5420e-02, 5.4443e-02, 5.3467e-02, ..., -1.3733e-04, + -6.7234e-05, -2.1935e-04], + [ 5.4688e-02, 5.2979e-02, 5.1025e-02, ..., 2.7275e-04, + 3.8528e-04, 4.1771e-04], + [-4.5654e-02, -4.4434e-02, -4.3213e-02, ..., -3.8445e-06, + 5.1498e-05, -5.6744e-05], + ..., + [ 2.8809e-02, 2.7954e-02, 2.7466e-02, ..., 1.0109e-04, + 3.3140e-05, 6.2943e-05], + [ 5.9814e-02, 5.8594e-02, 5.7861e-02, ..., 4.6921e-04, + 5.1117e-04, 5.0354e-04], + [-4.7607e-02, -4.6631e-02, -4.5654e-02, ..., -2.9755e-04, + -3.4142e-04, -3.0327e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0041, 0.0135, -0.0018, ..., -0.0096, -0.0148, 0.0106], + [-0.0134, 0.0061, 0.0108, ..., -0.0055, -0.0068, 0.0040], + [-0.0006, 0.0032, 0.0011, ..., -0.0012, -0.0017, 0.0017], + ..., + [ 0.0178, 0.0061, -0.0118, ..., 0.0077, 0.0155, 0.0041], + [-0.0043, 0.0393, 0.0046, ..., -0.0013, 0.0275, -0.0136], + [-0.0101, 0.0148, 0.0110, ..., -0.0106, -0.0070, 0.0193]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_mlp.1.weight': tensor([[-0.0518, -0.0491, -0.0476, ..., 0.0030, 0.0032, 0.0032], + [-0.0063, 0.0039, 0.0039, ..., 0.0007, 0.0008, 0.0007], + [ 0.0645, 0.0625, 0.0623, ..., -0.0016, -0.0016, -0.0016], + ..., + [ 0.0588, 0.0574, 0.0566, ..., -0.0017, -0.0017, -0.0015], + [ 0.0151, 0.0139, 0.0133, ..., -0.0010, -0.0009, -0.0010], + [-0.0598, -0.0586, -0.0576, ..., 0.0021, 0.0021, 0.0020]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_mlp.2.weight': tensor([[-0.0337, -0.0108, 0.0143, ..., 0.0069, 0.0125, -0.0147], + [ 0.0018, -0.0047, -0.0014, ..., -0.0011, 0.0047, 0.0011], + [ 0.0029, -0.0035, -0.0018, ..., -0.0018, 0.0041, 0.0018], + ..., + [ 0.0403, -0.0041, -0.0137, ..., -0.0083, 0.0041, 0.0143], + [ 0.0270, 0.0325, 0.0256, ..., 0.0282, -0.0325, -0.0258], + [ 0.0095, -0.0148, -0.0086, ..., -0.0112, 0.0151, 0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.after_proj.weight': tensor([[-0.0071, -0.0042, 0.0067, ..., -0.0049, 0.0068, -0.0044], + [ 0.0003, 0.0037, 0.0046, ..., -0.0112, -0.0070, -0.0019], + [-0.0046, -0.0081, 0.0173, ..., 0.0039, 0.0036, 0.0035], + ..., + [-0.0026, 0.0045, 0.0090, ..., -0.0181, 0.0084, 0.0073], + [ 0.0036, -0.0075, -0.0081, ..., 0.0049, -0.0157, 0.0065], + [-0.0063, -0.0060, 0.0150, ..., -0.0077, 0.0002, 0.0079]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.after_proj.bias': tensor([-0.0017, -0.0015, 0.0014, ..., -0.0042, 0.0008, 0.0029], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_proj.weight': tensor([[-0.0034, 0.0080, 0.0164, ..., -0.0070, 0.0063, -0.0270], + [-0.0066, 0.0023, -0.0067, ..., -0.0078, -0.0089, -0.0082], + [ 0.0114, 0.0062, 0.0042, ..., -0.0023, -0.0038, -0.0038], + ..., + [ 0.0045, -0.0076, -0.0040, ..., -0.0042, 0.0046, 0.0069], + [-0.0044, -0.0320, -0.0098, ..., -0.0381, -0.0189, -0.0298], + [-0.0075, -0.0038, -0.0040, ..., 0.0037, -0.0047, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_norm.weight': tensor([ 1.2656e+00, 1.6562e+00, 1.5078e+00, 1.7188e+00, 1.0498e-01, + 6.9824e-02, 4.7607e-02, 1.0192e-05, -1.2757e-30, -1.5564e-03, + 2.6978e-02, 1.0156e-01, -2.2095e-02, -2.9297e-02, 9.0790e-04, + 3.2616e-04, 6.0059e-02, 9.4922e-01, 9.9609e-01, 1.4267e-03, + 6.1035e-05, 1.1953e+00, 1.6406e+00, 1.6719e+00, 6.7969e-01, + 6.2891e-01, 1.5703e+00, -3.4332e-03, -1.0071e-02, 2.0146e-05, + 4.3869e-04, -5.0354e-03, 2.3682e-02, 1.0938e+00, 5.4932e-03, + 1.0781e+00, 1.5068e-04, 4.7207e-05, 1.3123e-02, -1.6327e-03, + 8.5449e-03, -3.2187e-05, 1.2054e-03, 1.6953e+00, 9.8438e-01, + 1.4062e+00, 1.6406e+00, 1.8516e+00, 1.0596e-01, 4.9973e-04, + 3.0029e-02, 2.2168e-01, 1.0859e+00, 8.2520e-02, 2.1820e-03, + 1.3086e-01, 5.6458e-04, 9.3750e-02, 7.4863e-05, -1.1902e-02, + 1.4267e-03, 3.8300e-03, -3.3722e-03, 1.1406e+00, 1.8828e+00, + 7.9297e-01, 9.5703e-01, 1.5312e+00, 4.0771e-02, -1.2390e-02, + 1.2085e-02, -1.9908e-05, 2.5761e-30, 1.1658e-02, 1.7822e-02, + 8.9722e-03, 6.0730e-03, 4.0588e-03, -2.5269e-02, 1.8066e-02, + 5.7861e-02, -2.6245e-03, 1.4587e-02, 2.7618e-03, 3.3379e-04, + 4.1504e-03, 5.5469e-01, 4.2578e-01, 1.4688e+00, 1.6406e+00, + 4.7656e-01, 5.7068e-03, -7.9727e-04, 2.9755e-04, -4.3335e-03, + 1.0300e-03, 3.4912e-02, 6.8848e-02, 5.2795e-03, 5.2002e-02, + 2.6321e-04, 4.1485e-05, 1.2891e-01, -2.6321e-04, 6.5918e-03, + -1.4484e-05, 1.3638e-04, 4.1797e-01, 1.5000e+00, 1.4062e+00, + 5.9375e-01, 9.2969e-01, 7.3242e-02, -2.4109e-03, 2.0020e-02, + 1.1641e+00, 5.0049e-02, 1.0859e+00, -1.5736e-05, 2.3651e-03, + 3.9307e-02, 4.5410e-02, 1.4210e-04, 1.9302e-03, -1.0872e-04, + -3.6011e-03, 9.4141e-01, -2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.self_attn.k_proj.weight': tensor([[-4.2725e-02, 1.7700e-03, -2.9449e-03, ..., -3.2959e-03, + -3.9101e-05, -1.3123e-02], + [-3.5156e-02, -1.2817e-02, -1.3123e-02, ..., 1.0071e-02, + -1.1230e-02, 2.4414e-02], + [ 4.5410e-02, 4.3030e-03, 3.8910e-03, ..., -1.2146e-02, + -4.1809e-03, -2.7100e-02], + ..., + [-7.1411e-03, 1.5259e-02, 5.7602e-04, ..., -5.2795e-03, + -6.2561e-03, -2.8076e-03], + [ 1.1597e-02, -8.8501e-03, -2.4261e-03, ..., -4.0894e-03, + 1.0986e-02, 3.2349e-03], + [ 2.0294e-03, 5.9814e-03, -1.2970e-03, ..., -5.9509e-03, + -6.1035e-04, 6.5308e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.k_norm.weight': tensor([ 7.4609e-01, 1.5938e+00, 1.5234e+00, 1.6562e+00, 1.0742e-01, + 7.5195e-02, 4.6387e-02, -1.3113e-05, 8.1351e-31, -2.1210e-03, + 2.6611e-02, 1.0205e-01, -2.6245e-02, 5.2979e-02, 1.1597e-03, + 1.4114e-04, 4.0588e-03, 9.4922e-01, 9.9609e-01, 1.3275e-03, + 1.0681e-04, 1.1953e+00, 1.4531e+00, 1.5469e+00, 3.5400e-02, + 1.4355e-01, 1.5938e+00, -5.7373e-03, 9.6436e-03, -9.6798e-05, + -1.5640e-03, 5.0964e-03, -2.0630e-02, 1.0938e+00, -2.0142e-03, + 1.0703e+00, 1.6975e-04, 4.4584e-05, -1.5076e-02, -1.7395e-03, + 8.5449e-03, -7.4387e-05, 1.2054e-03, 1.4609e+00, 4.4727e-01, + 1.3750e+00, 1.6953e+00, 1.8672e+00, -5.3955e-02, 1.8387e-03, + 1.0681e-02, 1.5625e-02, 1.0859e+00, 7.6172e-02, 2.1820e-03, + 1.3086e-01, -9.0122e-05, 9.3750e-02, 7.0095e-05, 7.6660e-02, + -1.4191e-03, 1.8311e-02, -1.9653e-02, 1.1406e+00, 1.9688e+00, + 1.1172e+00, 8.6328e-01, 1.5781e+00, 2.7710e-02, -5.7983e-03, + -1.3916e-02, 1.5974e-05, 1.9352e-30, 1.2024e-02, 3.1006e-02, + 5.0781e-02, 6.3782e-03, 2.6611e-02, -3.7842e-02, -1.8066e-02, + -1.8555e-02, 6.0730e-03, 9.0332e-03, 2.6550e-03, 3.2425e-04, + 8.6670e-03, 1.1172e+00, 1.0000e+00, 1.5625e+00, 1.7031e+00, + 1.4844e-01, 5.7678e-03, -3.0212e-03, -2.6131e-04, -3.6163e-03, + -6.2943e-04, 1.9287e-02, 9.0332e-03, 4.6387e-03, -8.9722e-03, + 2.7084e-04, -1.6332e-05, 1.2793e-01, -2.6321e-04, 6.5613e-03, + 2.1100e-05, 1.3256e-04, 1.1719e+00, 1.6406e+00, 1.4141e+00, + 1.1768e-01, 9.8438e-01, 8.7280e-03, 4.0894e-03, 5.0354e-03, + 1.1641e+00, -2.9755e-03, 1.0938e+00, -7.2002e-05, -1.8158e-03, + 3.9307e-02, 4.5166e-02, 1.3924e-04, 4.3457e-02, -1.1730e-04, + 1.5015e-02, 9.4141e-01, -1.5503e-02], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.self_attn.v_proj.weight': tensor([[-0.0017, 0.0042, 0.0115, ..., 0.0542, -0.0092, 0.0073], + [ 0.0060, 0.0214, -0.0223, ..., -0.0248, -0.0010, -0.0054], + [-0.0009, 0.0156, -0.0039, ..., 0.0126, -0.0044, 0.0096], + ..., + [ 0.0026, -0.0039, -0.0148, ..., 0.0102, 0.0055, -0.0009], + [ 0.0059, 0.0243, 0.0019, ..., 0.0025, 0.0025, 0.0002], + [-0.0074, -0.0142, 0.0027, ..., 0.0175, -0.0077, -0.0128]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.output_proj.weight': tensor([[ 0.0179, 0.0205, -0.0031, ..., 0.0021, 0.0260, -0.0159], + [ 0.0096, -0.0099, -0.0095, ..., -0.0045, -0.0166, 0.0025], + [ 0.0078, 0.0236, -0.0091, ..., 0.0062, -0.0017, -0.0050], + ..., + [-0.0287, -0.0366, -0.0045, ..., -0.0045, -0.0148, 0.0040], + [ 0.0026, 0.0087, -0.0182, ..., -0.0103, 0.0327, -0.0085], + [-0.0432, -0.0049, 0.0156, ..., 0.0238, -0.0083, -0.0036]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_proj.weight': tensor([[ 0.0049, 0.0067, 0.0038, ..., -0.0072, -0.0022, 0.0223], + [ 0.0061, -0.0040, -0.0039, ..., -0.0063, 0.0070, 0.0106], + [ 0.0023, 0.0029, -0.0030, ..., 0.0027, 0.0157, 0.0051], + ..., + [ 0.0076, -0.0088, -0.0059, ..., 0.0037, -0.0072, -0.0011], + [ 0.0017, -0.0032, -0.0029, ..., 0.0056, -0.0051, -0.0022], + [-0.0040, -0.0054, 0.0039, ..., -0.0080, 0.0072, 0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_norm.weight': tensor([0.1089, 0.0232, 0.4688, 0.0752, 0.1816, 0.0356, 0.1494, 0.1260, 0.5234, + 0.0388, 0.0630, 0.0148, 0.0630, 0.0762, 0.0508, 0.2021, 0.4883, 0.0260, + 0.7383, 0.0410, 0.0311, 0.1641, 0.0289, 0.0806, 0.0557, 0.4199, 0.5156, + 0.2754, 0.0713, 0.0454, 0.0233, 0.2451, 0.7031, 0.2617, 0.0415, 0.0344, + 0.0190, 0.1025, 0.0284, 0.0422, 0.0586, 0.1318, 0.0247, 0.0344, 0.2246, + 0.0425, 0.4043, 0.0304, 0.0540, 0.0095, 0.0259, 0.4141, 0.4902, 0.3867, + 0.0962, 0.6133, 0.5781, 0.0337, 0.1494, 0.0249, 0.0898, 0.4902, 0.0215, + 0.5547, 0.1699, 0.4180, 0.1040, 0.0488, 0.0723, 0.4941, 0.0262, 0.4902, + 0.0198, 0.0447, 0.1436, 0.0850, 0.0280, 0.1128, 0.6445, 0.0308, 0.0908, + 0.0874, 0.0417, 0.0381, 0.4297, 0.0723, 0.0515, 0.0874, 0.0211, 0.1787, + 0.0398, 0.1924, 0.0403, 0.0669, 0.0442, 0.1133, 0.0052, 0.3691, 0.5234, + 0.0471, 0.0581, 0.3301, 0.0347, 0.1270, 0.0352, 0.2656, 0.0214, 0.5664, + 0.5664, 0.0391, 0.0176, 0.7500, 0.0310, 0.0139, 0.5039, 0.2217, 0.3145, + 0.7109, 0.0332, 0.0913, 0.0884, 0.5391, 0.6016, 0.0146, 0.0248, 0.2041, + 0.0415, 0.0378], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_proj.weight': tensor([[-1.7548e-03, 1.8158e-03, 3.5706e-03, ..., -3.2349e-03, + 1.1353e-02, -1.2512e-02], + [ 5.9204e-03, -6.7711e-05, -2.9602e-03, ..., -3.0518e-04, + -1.3275e-03, 6.3171e-03], + [-1.1292e-02, 7.6904e-03, -1.3245e-02, ..., 6.8970e-03, + 9.8267e-03, 1.7853e-03], + ..., + [-1.6479e-03, 3.1128e-03, -5.8899e-03, ..., -3.5095e-03, + -1.3306e-02, 9.3384e-03], + [-1.2451e-02, -1.8539e-03, -8.9111e-03, ..., 7.5684e-03, + -9.2163e-03, -2.7275e-04], + [ 1.4893e-02, -1.4038e-03, 5.0049e-03, ..., -1.0071e-02, + 4.8523e-03, -3.5400e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_norm.weight': tensor([0.1084, 0.0232, 0.4688, 0.0752, 0.1816, 0.0361, 0.1494, 0.1260, 0.5234, + 0.0393, 0.0640, 0.0149, 0.0630, 0.0762, 0.0510, 0.2041, 0.4883, 0.0262, + 0.7344, 0.0410, 0.0309, 0.1641, 0.0291, 0.0801, 0.0557, 0.4160, 0.5156, + 0.2734, 0.0713, 0.0452, 0.0242, 0.2471, 0.7031, 0.2617, 0.0422, 0.0344, + 0.0188, 0.1025, 0.0287, 0.0417, 0.0583, 0.1328, 0.0251, 0.0349, 0.2246, + 0.0425, 0.4043, 0.0306, 0.0542, 0.0092, 0.0262, 0.4141, 0.4902, 0.3867, + 0.0962, 0.6133, 0.5820, 0.0334, 0.1494, 0.0253, 0.0898, 0.4902, 0.0219, + 0.5547, 0.1689, 0.4180, 0.1040, 0.0488, 0.0723, 0.4961, 0.0261, 0.4902, + 0.0195, 0.0444, 0.1436, 0.0845, 0.0280, 0.1138, 0.6445, 0.0305, 0.0898, + 0.0874, 0.0417, 0.0383, 0.4297, 0.0728, 0.0505, 0.0869, 0.0214, 0.1807, + 0.0408, 0.1914, 0.0400, 0.0674, 0.0447, 0.1143, 0.0052, 0.3691, 0.5234, + 0.0464, 0.0588, 0.3301, 0.0347, 0.1270, 0.0352, 0.2656, 0.0211, 0.5664, + 0.5664, 0.0391, 0.0177, 0.7500, 0.0305, 0.0141, 0.5078, 0.2217, 0.3145, + 0.7070, 0.0325, 0.0898, 0.0884, 0.5391, 0.6016, 0.0145, 0.0248, 0.2041, + 0.0427, 0.0381], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.v_proj.weight': tensor([[ 0.0099, -0.0024, -0.0236, ..., -0.0018, -0.0025, -0.0165], + [-0.0064, 0.0107, 0.0040, ..., 0.0040, 0.0023, -0.0250], + [ 0.0190, 0.0034, -0.0056, ..., 0.0033, -0.0052, -0.0396], + ..., + [ 0.0025, 0.0031, 0.0190, ..., -0.0087, -0.0259, 0.0049], + [-0.0332, -0.0007, -0.0219, ..., -0.0072, 0.0103, -0.0148], + [-0.0070, -0.0017, -0.0017, ..., -0.0082, 0.0104, 0.0071]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.output_proj.weight': tensor([[-0.0128, -0.0076, -0.0082, ..., -0.0063, 0.0133, 0.0111], + [-0.0248, -0.0042, -0.0100, ..., 0.0048, -0.0138, -0.0304], + [ 0.0011, 0.0024, 0.0064, ..., 0.0029, -0.0026, -0.0045], + ..., + [ 0.0141, -0.0109, 0.0123, ..., 0.0160, -0.0127, -0.0093], + [-0.0161, -0.0079, 0.0060, ..., 0.0239, -0.0284, 0.0334], + [-0.0165, -0.0115, 0.0254, ..., -0.0051, 0.0259, -0.0051]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_img.weight': tensor([[ 1.6235e-02, -4.2480e-02, -2.7954e-02, ..., 2.3682e-02, + -4.1748e-02, -4.6631e-02], + [-1.0925e-02, 2.9297e-02, 2.5635e-02, ..., -2.3315e-02, + -1.9775e-02, -2.3193e-02], + [-1.2146e-02, 4.9438e-03, -6.5918e-03, ..., 2.6611e-02, + -2.7832e-02, 7.5989e-03], + ..., + [-2.1729e-02, -4.1016e-02, 4.8828e-02, ..., 3.8862e-05, + -2.3956e-03, -2.5513e-02], + [ 9.9182e-04, 6.6833e-03, 3.8300e-03, ..., 3.0029e-02, + 8.4229e-03, 2.0508e-02], + [-2.1362e-02, 2.1729e-02, 5.6458e-03, ..., -1.3916e-02, + -1.4282e-02, -4.1992e-02]], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, 0.0013, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0024, 0.0293, ..., 0.0067, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0139, 0.0126, -0.0092], + ..., + [-0.0032, 0.0320, -0.0233, ..., -0.0062, -0.0028, 0.0188], + [ 0.0135, -0.0352, 0.0267, ..., -0.0305, -0.0039, 0.0018], + [-0.0205, 0.0140, 0.0039, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, 0.0005, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0056, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0015], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, 0.0013, 0.0012], + [-0.0141, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0023, ..., -0.0103, 0.0071, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.mlp.layer1.weight': tensor([[ 0.0104, 0.0049, -0.0160, ..., 0.0096, 0.0089, -0.0065], + [-0.0109, -0.0065, 0.0049, ..., -0.0120, -0.0062, -0.0079], + [-0.0034, 0.0359, -0.0165, ..., 0.0284, 0.0075, 0.0089], + ..., + [-0.0074, 0.0103, 0.0645, ..., -0.0041, 0.0085, -0.0117], + [-0.0038, 0.0049, 0.0042, ..., -0.0151, 0.0087, -0.0129], + [ 0.0055, 0.0066, -0.0072, ..., -0.0146, -0.0156, -0.0186]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.mlp.layer2.weight': tensor([[ 0.0111, -0.0063, 0.0005, ..., -0.0266, -0.0067, 0.0039], + [-0.0112, 0.0051, -0.0106, ..., -0.0027, 0.0101, -0.0044], + [ 0.0006, -0.0107, 0.0055, ..., 0.0058, 0.0098, -0.0042], + ..., + [ 0.0040, 0.0040, 0.0217, ..., -0.0044, 0.0064, -0.0055], + [ 0.0053, 0.0029, 0.0039, ..., 0.0110, 0.0036, 0.0095], + [-0.0168, -0.0084, 0.0029, ..., 0.0116, 0.0139, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0413, -0.0408, -0.0405, ..., 0.0009, 0.0007, 0.0008], + [-0.0369, -0.0366, -0.0356, ..., 0.0026, 0.0027, 0.0027], + [-0.0334, -0.0332, -0.0320, ..., 0.0021, 0.0022, 0.0022], + ..., + [ 0.0244, 0.0234, 0.0225, ..., -0.0037, -0.0038, -0.0037], + [ 0.0134, 0.0123, 0.0115, ..., -0.0032, -0.0031, -0.0031], + [-0.0344, -0.0337, -0.0330, ..., 0.0022, 0.0023, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0031, 0.0044, ..., 0.0063, 0.0090, 0.0045], + [-0.0028, -0.0012, -0.0015, ..., -0.0019, -0.0014, -0.0011], + [ 0.0024, -0.0069, -0.0049, ..., 0.0032, 0.0033, -0.0058], + ..., + [-0.0237, -0.0811, -0.0635, ..., 0.0327, 0.0332, -0.0723], + [-0.0161, -0.0140, -0.0139, ..., -0.0115, -0.0139, -0.0131], + [ 0.0098, 0.0067, 0.0069, ..., 0.0092, 0.0121, 0.0064]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_cross_attn.1.weight': tensor([[-7.9102e-02, -7.6660e-02, -7.5195e-02, ..., -9.5844e-05, + -1.1110e-04, -1.0157e-04], + [ 3.0273e-02, 2.9785e-02, 2.9297e-02, ..., 2.5481e-06, + 2.8610e-05, 1.4877e-04], + [ 2.5391e-02, 2.4658e-02, 2.3682e-02, ..., 2.4796e-04, + 2.7084e-04, 1.7738e-04], + ..., + [ 5.1514e-02, 5.0293e-02, 4.9805e-02, ..., -1.8692e-04, + -1.8120e-04, -1.1349e-04], + [ 5.5176e-02, 5.3711e-02, 5.2246e-02, ..., -1.8311e-04, + -1.2207e-04, -1.7071e-04], + [-1.6235e-02, -1.5747e-02, -1.5503e-02, ..., -5.0306e-05, + 1.0109e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0053, -0.0033, ..., 0.0281, 0.0317, 0.0069], + [ 0.0040, -0.0044, 0.0056, ..., -0.0009, 0.0025, -0.0019], + [ 0.0025, -0.0038, 0.0039, ..., -0.0026, -0.0026, -0.0044], + ..., + [-0.0069, -0.0052, 0.0403, ..., 0.0016, 0.0123, -0.0041], + [-0.0302, 0.0184, 0.0298, ..., 0.0115, 0.0130, -0.0046], + [ 0.0261, -0.0206, 0.0040, ..., 0.0048, 0.0107, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_mlp.1.weight': tensor([[-7.6172e-02, -7.3730e-02, -7.2754e-02, ..., 8.7261e-05, + 2.0409e-04, 1.5354e-04], + [ 5.8105e-02, 5.5420e-02, 5.5176e-02, ..., 3.8719e-04, + 4.6921e-04, 4.5776e-04], + [-9.2773e-02, -9.1309e-02, -8.9844e-02, ..., -4.7874e-04, + -4.1199e-04, -4.4250e-04], + ..., + [-7.4707e-02, -7.2754e-02, -7.1777e-02, ..., 1.1444e-04, + 2.1648e-04, 1.1492e-04], + [-6.0791e-02, -5.9326e-02, -5.8594e-02, ..., 1.5736e-04, + 1.8477e-05, 2.9922e-05], + [ 7.2266e-02, 7.1289e-02, 7.0312e-02, ..., -9.0599e-05, + -1.1873e-04, -8.2016e-05]], dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_mlp.2.weight': tensor([[-4.9438e-03, -3.9978e-03, 2.5177e-03, ..., -7.4158e-03, + -1.1414e-02, 8.9111e-03], + [-3.3417e-03, -2.6550e-03, -1.6861e-03, ..., -2.7618e-03, + -2.5787e-03, 2.5787e-03], + [-6.4850e-05, -2.4414e-03, -1.7471e-03, ..., -3.5477e-04, + -1.5259e-04, 1.3161e-04], + ..., + [-4.6143e-02, -6.9275e-03, -5.6396e-02, ..., -3.4424e-02, + -1.2695e-02, 2.6489e-02], + [ 8.1177e-03, -2.7344e-02, 3.1738e-02, ..., 6.6528e-03, + 5.4016e-03, -6.5613e-03], + [ 9.4604e-03, 7.5989e-03, 4.9438e-03, ..., 8.5449e-03, + 4.1199e-03, -6.7444e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.after_proj.weight': tensor([[ 0.0049, 0.0079, -0.0066, ..., 0.0017, 0.0227, 0.0034], + [-0.0020, -0.0085, -0.0080, ..., 0.0134, -0.0012, -0.0102], + [-0.0118, 0.0030, -0.0039, ..., -0.0023, -0.0021, 0.0011], + ..., + [-0.0079, -0.0026, 0.0031, ..., -0.0070, 0.0021, -0.0082], + [-0.0040, 0.0076, 0.0044, ..., 0.0051, -0.0108, -0.0024], + [ 0.0119, 0.0110, 0.0015, ..., -0.0018, 0.0058, 0.0067]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.after_proj.bias': tensor([ 0.0050, -0.0118, -0.0096, ..., -0.0070, 0.0018, 0.0004], + dtype=torch.bfloat16)} +[06-26 08:14:38|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:38|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 08:14:38|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 1 checkpoint. +[06-26 08:14:38|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/depth(626e6618-bfcd-4d9a-a077-1409e2ce353f) +[06-26 08:14:38|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 2 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/dea7737ca29dd8d9086413c6dc5724b8250a0bb4/general/depth/626e6618-bfcd-4d9a-a077-1409e2ce353f_ema_bf16.pt +[06-26 08:14:39|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0144, -0.0021, 0.0064], + [ 0.0036, 0.0057, -0.0327, ..., -0.0082, -0.0024, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0137, -0.0029, 0.0283], + ..., + [ 0.0010, 0.0125, 0.0248, ..., 0.0437, -0.0074, -0.0192], + [ 0.0277, 0.0146, 0.0036, ..., -0.0031, -0.0092, 0.0146], + [ 0.0437, 0.0366, 0.0085, ..., 0.0013, -0.0128, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0028, -0.0310, 0.0300, ..., -0.0071, -0.0042, -0.0339], + [-0.0103, -0.0048, 0.0115, ..., 0.0049, 0.0092, 0.0520], + [-0.0064, 0.0132, 0.0398, ..., -0.0097, -0.0179, -0.0070], + ..., + [-0.0046, 0.0175, -0.0094, ..., 0.0225, 0.0047, -0.0020], + [-0.0209, -0.0081, 0.0087, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0354, -0.0082, ..., 0.0123, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0016, ..., 0.0376, 0.0532, 0.0011], + [ 0.0175, -0.0347, 0.0015, ..., -0.0064, 0.0047, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0156, 0.0262, -0.0114], + [-0.0015, 0.0244, 0.0356, ..., 0.0018, 0.0170, 0.0112], + [ 0.0486, 0.0175, 0.0042, ..., 0.0046, -0.0021, 0.0008]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0118, ..., -0.0075, -0.0337, 0.0079], + [ 0.0491, 0.0413, 0.0251, ..., -0.0041, 0.0393, 0.0035], + [-0.0162, -0.0005, 0.0134, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0014, 0.0261, 0.0042, ..., -0.0038, 0.0347, 0.0036], + [-0.0249, 0.0067, 0.0024, ..., 0.0096, 0.0388, -0.0205], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0123, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0048, -0.0270, -0.0166, ..., 0.0118, -0.0298, 0.0532], + [ 0.0034, 0.0036, -0.0126, ..., -0.0161, 0.0152, 0.0308], + [ 0.0145, 0.0086, 0.0137, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0067, -0.0630, 0.0027, ..., 0.0072, -0.0183, 0.0217], + [ 0.0039, 0.0032, 0.0258, ..., -0.0486, -0.0349, -0.0039], + [ 0.0182, -0.0081, 0.0197, ..., 0.0151, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0053, -0.0042, -0.0005, ..., -0.0007, -0.0017, -0.0121], + [-0.0152, 0.0251, 0.0432, ..., -0.0051, 0.0039, -0.0038], + [-0.0149, -0.0203, -0.0245, ..., -0.0118, -0.0325, 0.0029], + ..., + [ 0.0043, 0.0008, 0.0417, ..., 0.0280, 0.0022, -0.0024], + [-0.0039, 0.0037, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0067, -0.0034, -0.0085]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0155, -0.0063, -0.0481, ..., 0.0449, 0.0212, 0.0128], + [ 0.0156, -0.0227, 0.0220, ..., -0.0219, 0.0005, 0.0081], + [-0.0156, -0.0098, 0.0430, ..., -0.0101, 0.0327, -0.0110], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0136], + [ 0.0021, 0.0166, -0.0061, ..., 0.0013, 0.0079, -0.0052], + [-0.0110, -0.0041, 0.0117, ..., 0.0047, -0.0034, -0.0077]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0035, -0.0386, -0.0120, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0075, 0.0286, ..., -0.0339, -0.0160, 0.0170], + [-0.0127, -0.0053, -0.0128, ..., 0.0214, 0.0111, -0.0044], + ..., + [-0.0153, -0.0172, -0.0137, ..., 0.0098, 0.0098, 0.0184], + [-0.0154, -0.0356, 0.0201, ..., -0.0069, -0.0117, 0.0332], + [-0.0396, -0.0121, -0.0090, ..., -0.0284, -0.0141, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0280, 0.0312, 0.0133, ..., -0.0195, 0.0114, -0.0153], + [ 0.0192, 0.0092, -0.0073, ..., 0.0022, -0.0127, -0.0198], + [-0.0249, 0.0109, -0.0137, ..., 0.0090, 0.0172, 0.0067], + ..., + [ 0.0466, -0.0045, 0.0007, ..., 0.0010, 0.0239, 0.0042], + [ 0.0547, -0.0194, -0.0108, ..., -0.0069, 0.0311, -0.0044], + [ 0.0051, 0.0104, -0.0095, ..., 0.0203, 0.0170, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0044, -0.0046, 0.0283, ..., -0.0109, -0.0045, -0.0182], + [ 0.0052, 0.0126, -0.0315, ..., -0.0457, -0.0069, 0.0109], + [ 0.0006, 0.0261, 0.0245, ..., 0.0046, 0.0084, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0155, 0.0058, -0.0077], + [ 0.0059, -0.0371, 0.0356, ..., -0.0082, 0.0410, -0.0178], + [ 0.0178, -0.0155, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0087, 0.0208, 0.0272, ..., 0.0137, 0.0156, 0.0082], + [-0.0082, -0.0023, -0.0041, ..., 0.0144, 0.0007, 0.0057], + [ 0.0046, 0.0010, 0.0240, ..., -0.0119, -0.0066, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0046, ..., 0.0248, -0.0106, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., 0.0020, -0.0295, 0.0400], + [-0.0118, 0.0135, 0.0305, ..., -0.0100, 0.0366, 0.0227]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0054, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0149, -0.0054, 0.0288, ..., 0.0136, 0.0583, 0.0044], + [-0.0281, 0.0154, -0.0386, ..., -0.0444, -0.0093, -0.0009], + ..., + [ 0.0035, 0.0157, 0.0217, ..., 0.0047, 0.0115, 0.0151], + [ 0.0051, 0.0195, -0.0283, ..., -0.0048, -0.0039, -0.0342], + [-0.0167, -0.0195, 0.0007, ..., -0.0048, 0.0165, 0.0132]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0140, -0.0146, 0.0459, ..., 0.0043, -0.0320, 0.0275], + [ 0.0034, 0.0061, 0.0272, ..., -0.0219, -0.0183, -0.0265], + [ 0.0219, 0.0216, -0.0201, ..., -0.0022, 0.0052, 0.0041], + ..., + [-0.0258, 0.0028, 0.0215, ..., 0.0017, 0.0236, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0060], + [-0.0037, -0.0085, -0.0128, ..., -0.0334, 0.0062, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0146, -0.0454, ..., -0.0126, -0.0131, 0.0400], + [ 0.0295, -0.0187, -0.0106, ..., -0.0283, 0.0040, -0.0186], + [ 0.0013, 0.0547, -0.0042, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0280, 0.0045, -0.0269, ..., -0.0090, -0.0287, 0.0150], + [ 0.0093, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0098, 0.0106, ..., -0.0075, 0.0703, -0.0059]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0094, -0.0315, ..., 0.0058, 0.0254, 0.0359], + [ 0.0087, -0.0048, 0.0028, ..., -0.0125, -0.0233, -0.0042], + [ 0.0070, 0.0054, -0.0121, ..., -0.0123, 0.0018, 0.0089], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0040], + [ 0.0339, -0.0059, 0.0223, ..., 0.0410, 0.0028, 0.0138], + [-0.0055, -0.0439, -0.0486, ..., 0.0140, -0.0130, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0002, 0.0139, -0.0042, ..., -0.0079, -0.0045, -0.0229], + [-0.0289, 0.0231, 0.0084, ..., -0.0140, 0.0056, 0.0222], + [ 0.0085, 0.0295, -0.0193, ..., -0.0164, 0.0310, -0.0262], + ..., + [ 0.0233, -0.0053, -0.0029, ..., 0.0051, 0.0278, 0.0031], + [-0.0104, -0.0018, 0.0140, ..., -0.0175, 0.0444, 0.0425], + [ 0.0054, -0.0084, 0.0096, ..., 0.0061, -0.0344, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[ 0.0027, 0.0126, 0.0188, ..., -0.0092, -0.0199, 0.0147], + [-0.0581, 0.0294, -0.0043, ..., -0.0240, 0.0058, 0.0302], + [ 0.0087, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [ 0.0002, -0.0292, 0.0537, ..., 0.0033, -0.0078, -0.0106], + [-0.0036, -0.0077, 0.0209, ..., 0.0043, 0.0093, 0.0187], + [-0.0087, -0.0349, -0.0173, ..., -0.0400, 0.0057, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0182, -0.0265, -0.0105, ..., 0.0099, -0.0132, -0.0405], + [ 0.0051, -0.0087, 0.0266, ..., -0.0400, 0.0060, 0.0391], + [-0.0016, 0.0359, 0.0284, ..., 0.0292, -0.0233, 0.0212], + ..., + [ 0.0026, 0.0121, 0.0132, ..., -0.0141, -0.0048, 0.0095], + [ 0.0299, -0.0167, -0.0042, ..., -0.0013, -0.0002, -0.0206], + [ 0.0072, -0.0045, -0.0042, ..., 0.0037, -0.0022, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, -0.0015, 0.0118, ..., -0.0029, -0.0208, -0.0074], + [-0.0096, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0232], + [ 0.0134, -0.0278, 0.0118, ..., 0.0396, -0.0079, -0.0069], + ..., + [-0.0001, -0.0018, -0.0311, ..., 0.0344, 0.0051, -0.0123], + [-0.0236, -0.0102, -0.0072, ..., 0.0332, -0.0114, -0.0179], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0137, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[-0.0014, 0.0166, 0.0036, ..., 0.0273, -0.0352, 0.0156], + [-0.0510, -0.0062, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0077, ..., -0.0388, -0.0019, 0.0398], + ..., + [-0.0160, -0.0195, -0.0027, ..., 0.0037, -0.0056, 0.0029], + [-0.0187, 0.0038, 0.0098, ..., 0.0287, -0.0125, -0.0095], + [-0.0181, -0.0032, 0.0247, ..., -0.0143, 0.0052, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0096, 0.0141, ..., -0.0112, 0.0026, -0.0253], + [-0.0167, 0.0330, -0.0171, ..., -0.0037, -0.0403, -0.0110], + [ 0.0084, 0.0347, 0.0275, ..., 0.0148, 0.0064, 0.0079], + ..., + [ 0.0272, -0.0168, -0.0106, ..., -0.0098, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0059, ..., -0.0032, -0.0018, -0.0243], + [-0.0075, 0.0581, 0.0057, ..., 0.0145, 0.0239, 0.0116]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0118, -0.0410, 0.0126, ..., -0.0254, -0.0052, -0.0173], + [ 0.0476, 0.0315, 0.0148, ..., -0.0150, -0.0066, -0.0012], + [-0.0089, -0.0269, 0.0083, ..., 0.0332, 0.0144, -0.0063], + ..., + [-0.0046, -0.0043, 0.0160, ..., -0.0177, 0.0168, -0.0126], + [ 0.0051, -0.0187, -0.0049, ..., 0.0248, -0.0148, -0.0068], + [-0.0142, -0.0024, 0.0148, ..., -0.0043, 0.0037, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0054, 0.0094, ..., 0.0255, -0.0253, -0.0135], + [-0.0018, -0.0143, -0.0160, ..., -0.0113, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0115, 0.0046, 0.0322, ..., -0.0183, -0.0010, 0.0016], + [ 0.0034, -0.0050, 0.0162, ..., 0.0043, 0.0255, 0.0063], + [ 0.0270, 0.0087, 0.0159, ..., 0.0189, 0.0148, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0066, ..., -0.0532, 0.0104, 0.0036], + [-0.0240, -0.0215, -0.0103, ..., -0.0059, 0.0098, -0.0168], + [ 0.0074, -0.0013, -0.0059, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0057, 0.0172, -0.0106, ..., -0.0300, -0.0045, 0.0435], + [ 0.0051, -0.0261, 0.0247, ..., 0.0192, -0.0308, 0.0061], + [-0.0167, 0.0123, 0.0106, ..., -0.0247, 0.0123, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0078, -0.0078, 0.0317, ..., 0.0200, -0.0104, -0.0376], + [ 0.0356, 0.0378, 0.0135, ..., 0.0325, 0.0391, 0.0173], + [-0.0244, -0.0264, 0.0371, ..., 0.0028, 0.0052, -0.0322], + ..., + [ 0.0058, 0.0107, -0.0040, ..., 0.0366, -0.0206, 0.0374], + [-0.0046, -0.0070, 0.0034, ..., -0.0093, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0026, ..., -0.0166, -0.0056, 0.0245]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0053, -0.0093, -0.0344, ..., -0.0076, 0.0347, -0.0176], + [ 0.0267, -0.0366, -0.0088, ..., 0.0217, 0.0349, -0.0043], + [-0.0097, -0.0286, 0.0074, ..., 0.0052, -0.0282, 0.0013], + ..., + [-0.0129, 0.0037, 0.0058, ..., -0.0086, 0.0165, 0.0076], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0129], + [-0.0140, 0.0043, -0.0170, ..., -0.0127, 0.0060, 0.0070]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0025, -0.0047, ..., -0.0483, 0.0374, 0.0093], + [-0.0576, 0.0116, 0.0374, ..., -0.0153, -0.0188, -0.0136], + ..., + [ 0.0260, -0.0007, -0.0077, ..., -0.0003, 0.0053, -0.0165], + [-0.0222, -0.0155, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0131, -0.0148, -0.0038, ..., 0.0018, -0.0234, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0101], + [ 0.0249, 0.0205, 0.0075, ..., -0.0061, -0.0089, -0.0105], + [-0.0220, 0.0044, 0.0047, ..., 0.0309, 0.0381, 0.0037], + ..., + [ 0.0200, -0.0033, 0.0171, ..., 0.0122, 0.0320, -0.0157], + [ 0.0349, 0.0120, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0075, 0.0256, 0.0050, ..., 0.0205, 0.0035, -0.0105]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0043, 0.0199, ..., -0.0050, 0.0276, -0.0308], + [ 0.0178, -0.0063, 0.0099, ..., 0.0108, -0.0398, 0.0300], + [ 0.0430, -0.0124, 0.0081, ..., -0.0009, -0.0066, -0.0055], + ..., + [ 0.0192, -0.0106, -0.0108, ..., -0.0025, 0.0330, 0.0142], + [-0.0124, -0.0542, -0.0205, ..., 0.0062, -0.0159, 0.0320], + [ 0.0303, 0.0054, -0.0425, ..., 0.0009, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0068, -0.0315, ..., 0.0525, 0.0034, -0.0349], + [-0.0044, -0.0070, -0.0156, ..., 0.0053, -0.0240, 0.0305], + [-0.0256, -0.0045, 0.0062, ..., -0.0459, -0.0148, -0.0165], + ..., + [ 0.0070, 0.0231, 0.0209, ..., 0.0077, -0.0038, 0.0461], + [-0.0067, -0.0085, 0.0004, ..., -0.0217, 0.0347, -0.0018], + [ 0.0057, 0.0234, 0.0056, ..., 0.0261, -0.0038, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0071, 0.0023], + [-0.0183, -0.0211, 0.0045, ..., -0.0145, 0.0143, -0.0166], + [-0.0075, -0.0160, 0.0053, ..., -0.0203, 0.0267, 0.0217], + ..., + [ 0.0308, 0.0096, 0.0075, ..., 0.0171, 0.0053, -0.0070], + [-0.0145, 0.0260, -0.0085, ..., 0.0077, 0.0102, -0.0166], + [ 0.0094, -0.0142, -0.0037, ..., 0.0051, 0.0069, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0070, 0.0130, -0.0076, ..., -0.0278, 0.0240, -0.0126], + [ 0.0035, 0.0187, -0.0160, ..., -0.0317, -0.0032, 0.0173], + [ 0.0062, 0.0310, -0.0092, ..., -0.0226, 0.0030, 0.0334], + ..., + [-0.0079, 0.0186, 0.0038, ..., 0.0228, 0.0069, 0.0065], + [ 0.0271, 0.0117, 0.0354, ..., -0.0076, 0.0198, 0.0145], + [ 0.0282, 0.0171, 0.0317, ..., -0.0111, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0114, -0.0265, -0.0110, ..., 0.0049, 0.0037, 0.0136], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0085], + [-0.0067, -0.0190, -0.0045, ..., 0.0166, -0.0135, 0.0138], + ..., + [-0.0200, -0.0012, 0.0061, ..., -0.0339, 0.0361, -0.0025], + [ 0.0037, -0.0250, 0.0153, ..., 0.0006, -0.0162, 0.0220], + [-0.0078, 0.0028, 0.0205, ..., -0.0064, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0053, 0.0240, -0.0217, ..., 0.0172, 0.0048, 0.0114], + [-0.0386, 0.0103, 0.0075, ..., -0.0042, 0.0322, 0.0086], + [-0.0427, 0.0040, 0.0035, ..., 0.0073, 0.0110, -0.0137], + ..., + [-0.0087, -0.0041, -0.0264, ..., -0.0586, -0.0630, -0.0132], + [ 0.0045, 0.0287, 0.0420, ..., -0.0311, -0.0090, -0.0289], + [-0.0107, -0.0610, 0.0162, ..., -0.0291, -0.0166, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0121, -0.0251, 0.0211, ..., 0.0199, -0.0084, -0.0112], + [ 0.0112, -0.0099, -0.0327, ..., 0.0447, 0.0076, 0.0074], + [ 0.0035, -0.0175, 0.0066, ..., -0.0062, -0.0049, -0.0065], + ..., + [ 0.0295, 0.0045, -0.0104, ..., 0.0093, -0.0160, -0.0288], + [ 0.0083, 0.0253, -0.0177, ..., 0.0094, 0.0074, 0.0080], + [-0.0129, 0.0027, 0.0137, ..., -0.0128, 0.0277, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0131, 0.0047, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0068, 0.0234, 0.0065, ..., 0.0464, 0.0064, 0.0309], + ..., + [-0.0117, -0.0175, 0.0327, ..., 0.0132, 0.0044, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, 0.0059, -0.0444, ..., 0.0060, -0.0283, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0082, -0.0099, 0.0146, ..., -0.0742, -0.0064, 0.0161], + [-0.0076, -0.0131, -0.0075, ..., 0.0064, 0.0026, 0.0238], + [ 0.0189, 0.0221, -0.0088, ..., 0.0050, 0.0408, 0.0052], + ..., + [ 0.0059, 0.0247, 0.0396, ..., 0.0064, -0.0304, -0.0183], + [ 0.0299, -0.0114, -0.0417, ..., -0.0048, -0.0376, -0.0044], + [-0.0210, 0.0168, -0.0026, ..., 0.0049, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0069, 0.0242, 0.0062, ..., -0.0245, 0.0078, 0.0002], + [ 0.0190, -0.0125, 0.0332, ..., -0.0146, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0068, ..., -0.0059, -0.0092, 0.0182], + ..., + [-0.0430, -0.0403, -0.0081, ..., 0.0135, -0.0140, 0.0046], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0094, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0042, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0092, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0067, ..., 0.0101, -0.0223, 0.0062], + ..., + [ 0.0192, 0.0292, 0.0059, ..., 0.0176, -0.0197, 0.0046], + [ 0.0265, 0.0063, 0.0054, ..., 0.0286, -0.0021, -0.0267], + [-0.0184, -0.0088, -0.0048, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0084, -0.0159, 0.0082, ..., 0.0198, 0.0056, 0.0277], + [ 0.0172, 0.0041, -0.0057, ..., 0.0048, -0.0049, 0.0057], + [ 0.0036, 0.0184, 0.0113, ..., 0.0099, -0.0056, -0.0074], + ..., + [ 0.0048, -0.0111, 0.0035, ..., -0.0096, -0.0053, 0.0122], + [-0.0427, -0.0391, -0.0192, ..., 0.0091, -0.0238, -0.0078], + [ 0.0055, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0084, -0.0161, 0.0123], + [-0.0081, 0.0060, 0.0312, ..., 0.0053, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0071, -0.0014, 0.0278], + ..., + [ 0.0170, -0.0085, 0.0471, ..., 0.0067, 0.0078, -0.0073], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0076, -0.0203], + [ 0.0053, 0.0048, -0.0081, ..., -0.0095, -0.0086, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[-0.0058, -0.0156, 0.0284, ..., -0.0104, 0.0135, -0.0044], + [ 0.0152, 0.0332, 0.0337, ..., 0.0223, -0.0165, -0.0044], + [ 0.0055, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [-0.0056, -0.0019, 0.0280, ..., -0.0076, -0.0078, 0.0140], + [-0.0083, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0091, ..., 0.0067, 0.0167, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0042, ..., -0.0238, -0.0103, 0.0119], + [-0.0250, -0.0162, 0.0034, ..., -0.0081, 0.0002, -0.0369], + [ 0.0051, 0.0141, 0.0079, ..., -0.0452, -0.0038, 0.0120], + ..., + [ 0.0148, -0.0141, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0101, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0142, 0.0041, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[ 0.0028, -0.0327, -0.0190, ..., 0.0064, -0.0052, -0.0209], + [ 0.0713, 0.0152, -0.0052, ..., -0.0049, 0.0186, 0.0027], + [-0.0227, -0.0092, 0.0188, ..., -0.0122, 0.0193, 0.0053], + ..., + [ 0.0116, 0.0464, 0.0085, ..., 0.0248, 0.0168, 0.0047], + [ 0.0381, -0.0038, 0.0071, ..., -0.0031, -0.0107, -0.0054], + [ 0.0292, 0.0057, 0.0209, ..., 0.0056, 0.0234, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0056, 0.0212, ..., 0.0305, 0.0137, 0.0105], + [-0.0078, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0045], + ..., + [ 0.0084, -0.0275, 0.0225, ..., 0.0043, -0.0189, -0.0107], + [-0.0128, 0.0053, -0.0058, ..., -0.0079, 0.0064, 0.0211], + [ 0.0015, -0.0073, -0.0063, ..., -0.0109, 0.0264, 0.0059]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[-0.0049, -0.0075, -0.0108, ..., 0.0186, -0.0216, -0.0047], + [ 0.0048, 0.0123, -0.0231, ..., 0.0123, -0.0112, -0.0183], + [ 0.0107, 0.0103, -0.0057, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0125, 0.0154, ..., 0.0540, 0.0415, 0.0097], + [ 0.0025, 0.0087, -0.0137, ..., -0.0030, -0.0189, -0.0237], + [ 0.0059, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0076, 0.0359, ..., -0.0084, -0.0120, -0.0049], + [ 0.0190, -0.0059, 0.0298, ..., 0.0126, -0.0050, -0.0200], + [ 0.0112, -0.0110, 0.0177, ..., -0.0248, 0.0229, 0.0034], + ..., + [-0.0081, 0.0044, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0044, -0.0291, ..., 0.0206, 0.0304, -0.0085], + [-0.0080, 0.0245, -0.0068, ..., -0.0413, 0.0216, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0078, -0.0157, 0.0043, ..., -0.0052, 0.0228, -0.0339], + [-0.0713, -0.0153, -0.0168, ..., -0.0217, -0.0183, -0.0073], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, -0.0063], + ..., + [-0.0027, -0.0069, 0.0258, ..., -0.0310, 0.0091, -0.0361], + [ 0.0063, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [-0.0071, 0.0157, -0.0069, ..., 0.0075, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0077, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0106, ..., 0.0104, -0.0210, 0.0148], + [-0.0286, -0.0267, -0.0037, ..., 0.0129, -0.0059, 0.0087], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [-0.0062, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0129, 0.0206, ..., 0.0069, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0152, -0.0092, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0045, 0.0076, ..., -0.0046, 0.0110, 0.0598], + [-0.0062, 0.0267, -0.0053, ..., 0.0284, -0.0045, -0.0364], + ..., + [ 0.0045, 0.0078, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0017, -0.0087, ..., 0.0211, 0.0137, -0.0075], + [ 0.0101, 0.0236, -0.0206, ..., -0.0136, -0.0063, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0067, 0.0294, ..., 0.0040, 0.0270, 0.0065], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0066, -0.0168], + [ 0.0201, -0.0121, 0.0126, ..., -0.0381, -0.0247, -0.0059], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0078], + [-0.0194, 0.0094, -0.0092, ..., -0.0231, -0.0069, -0.0466], + [ 0.0327, 0.0284, -0.0044, ..., 0.0145, 0.0142, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, -0.0021, -0.0167, ..., -0.0037, 0.0170, 0.0107], + [ 0.0019, -0.0031, 0.0295, ..., -0.0172, 0.0040, 0.0135], + [ 0.0193, 0.0236, -0.0125, ..., 0.0038, -0.0369, 0.0239], + ..., + [ 0.0151, -0.0413, -0.0124, ..., -0.0354, 0.0065, -0.0320], + [-0.0266, -0.0063, -0.0327, ..., -0.0050, 0.0058, 0.0073], + [ 0.0298, 0.0356, -0.0192, ..., 0.0042, 0.0008, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0045, 0.0058, 0.0317, ..., 0.0062, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0071, -0.0096], + [-0.0166, 0.0064, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0041, 0.0260, ..., 0.0159, -0.0035, -0.0474], + [-0.0064, 0.0079, 0.0076, ..., 0.0140, 0.0557, 0.0052], + [-0.0203, 0.0273, -0.0040, ..., -0.0330, -0.0038, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-3.3203e-02, -1.2390e-02, -3.8757e-03, ..., 6.3782e-03, + -9.3384e-03, -2.7710e-02], + [ 4.9744e-03, -3.4637e-03, -5.5237e-03, ..., 6.5002e-03, + 1.4038e-02, -1.6113e-02], + [-4.6997e-03, 1.6846e-02, -7.6599e-03, ..., -9.1553e-03, + 1.2390e-02, 2.0447e-03], + ..., + [ 3.9795e-02, 4.5166e-02, 2.5146e-02, ..., -5.8289e-03, + -3.2959e-02, 4.1199e-03], + [-6.2012e-02, 3.5248e-03, 3.7109e-02, ..., 4.1016e-02, + 9.2506e-05, -1.7334e-02], + [ 1.2634e-02, 5.7678e-03, -2.6978e-02, ..., 4.9438e-03, + 3.8818e-02, -1.0376e-02]], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0013, 0.0245], + [-0.0449, 0.0019, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, 0.0062, ..., 0.0041, 0.0288, 0.0124], + [ 0.0330, 0.0038, 0.0060, ..., 0.0299, 0.0142, -0.0483], + [-0.0166, 0.0077, 0.0051, ..., -0.0062, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0045, -0.0025, 0.0527], + [-0.0002, 0.0403, 0.0112, ..., 0.0417, 0.0105, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0096, -0.0096, 0.0327], + [ 0.0142, 0.0111, -0.0069, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0124, 0.0281, ..., -0.0242, -0.0127, -0.0047]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0042], + [ 0.0063, -0.0045, -0.0286, ..., 0.0056, 0.0172, 0.0106], + [ 0.0393, 0.0400, 0.0039, ..., 0.0025, 0.0082, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0107, 0.0097, 0.0150], + [-0.0057, 0.0039, 0.0077, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0078, ..., 0.0092, -0.0114, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0071, 0.0041, ..., -0.0220, -0.0145, -0.0013], + [-0.0166, 0.0189, -0.0047, ..., 0.0096, 0.0020, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0054, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0057, ..., 0.0047, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0082, 0.0071, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0122, 0.0289, 0.0125, ..., 0.0135, 0.0070, -0.0104], + [ 0.0188, -0.0197, 0.0010, ..., 0.0284, -0.0342, -0.0057], + [-0.0525, -0.0215, 0.0154, ..., -0.0270, 0.0081, 0.0254], + ..., + [-0.0069, 0.0087, 0.0281, ..., 0.0129, -0.0064, -0.0048], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0045, 0.0038], + [-0.0273, -0.0747, 0.0208, ..., -0.0044, -0.0044, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0048, 0.0150, 0.0334, ..., 0.0205, -0.0273, -0.0037], + [-0.0165, 0.0111, -0.0135, ..., 0.0106, -0.0071, 0.0219], + [ 0.0063, 0.0227, -0.0054, ..., 0.0128, -0.0077, 0.0150], + ..., + [-0.0347, -0.0148, 0.0053, ..., -0.0046, -0.0042, -0.0094], + [-0.0077, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0069, -0.0500, -0.0232, ..., 0.0320, 0.0057, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0056, -0.0364], + [ 0.0220, -0.0026, -0.0112, ..., -0.0052, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, 0.0039, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0064, 0.0192, -0.0393], + [-0.0102, -0.0244, 0.0275, ..., 0.0029, 0.0042, 0.0081], + [-0.0170, -0.0427, 0.0140, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0040, -0.0378, -0.0359, ..., 0.0552, -0.0062, -0.0160], + [ 0.0039, 0.0143, 0.0134, ..., 0.0049, -0.0153, 0.0354], + [ 0.0095, -0.0038, 0.0065, ..., -0.0046, -0.0249, -0.0136], + ..., + [ 0.0081, -0.0124, 0.0050, ..., -0.0095, -0.0085, 0.0047], + [-0.0051, 0.0439, 0.0166, ..., 0.0248, 0.0131, -0.0187], + [ 0.0175, 0.0059, 0.0041, ..., 0.0410, -0.0096, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0104, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0141, -0.0050, ..., 0.0052, -0.0244, 0.0356], + [ 0.0063, 0.0165, -0.0200, ..., -0.0019, 0.0237, -0.0078], + ..., + [-0.0197, -0.0049, 0.0386, ..., 0.0181, 0.0112, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, 0.0041, -0.0374], + [-0.0086, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0066, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0135, 0.0114, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, 0.0045, -0.0041, ..., 0.0050, -0.0055, 0.0337], + ..., + [ 0.0422, 0.0033, 0.0352, ..., -0.0304, -0.0049, -0.0204], + [ 0.0302, 0.0156, 0.0099, ..., -0.0078, -0.0254, -0.0245], + [-0.0040, -0.0060, -0.0039, ..., 0.0037, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0052, 0.0266, ..., 0.0154, 0.0187, 0.0047], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0051, ..., -0.0172, -0.0091, 0.0231], + ..., + [-0.0295, -0.0137, 0.0038, ..., 0.0043, -0.0044, 0.0099], + [-0.0049, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0078], + [-0.0040, -0.0048, -0.0259, ..., -0.0039, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0079, 0.0078, -0.0140, ..., 0.0242, -0.0070, -0.0386], + [-0.0140, 0.0079, -0.0097, ..., -0.0369, -0.0156, 0.0157], + [-0.0267, 0.0078, -0.0057, ..., -0.0126, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0007, 0.0144, ..., 0.0115, 0.0042, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., -0.0046, 0.0042, -0.0137], + [-0.0219, -0.0396, 0.0041, ..., -0.0058, -0.0018, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0034, 0.0168, 0.0194, ..., 0.0349, -0.0420, 0.0017], + [ 0.0042, 0.0273, 0.0145, ..., -0.0184, -0.0039, 0.0154], + [ 0.0145, -0.0036, -0.0072, ..., 0.0245, 0.0359, 0.0059], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0046, 0.0033, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0070, -0.0312, -0.0457], + [ 0.0050, 0.0054, -0.0053, ..., 0.0049, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0015, 0.0123, -0.0029, ..., 0.0124, -0.0038, 0.0066], + [-0.0277, -0.0036, -0.0181, ..., -0.0027, -0.0034, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0042, 0.0513, -0.0110], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0111], + [-0.0282, -0.0053, 0.0121, ..., -0.0051, -0.0039, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0039, -0.0128, ..., 0.0140, 0.0308, 0.0058], + [ 0.0119, -0.0332, -0.0320, ..., -0.0078, -0.0140, -0.0029], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, -0.0028, -0.0311, ..., 0.0120, 0.0061, 0.0081], + [ 0.0048, 0.0198, 0.0142, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, -0.0046, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0091, -0.0036, 0.0056, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0018, ..., 0.0302, -0.0024, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0039, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0025, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0049, -0.0222, 0.0262], + ..., + [-0.0147, 0.0058, 0.0260, ..., 0.0051, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0057, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0076, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0113], + [-0.0278, 0.0044, -0.0374, ..., 0.0276, -0.0038, 0.0151], + [ 0.0089, 0.0237, -0.0067, ..., 0.0527, -0.0070, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0068, -0.0095], + [-0.0075, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0044, -0.0128, ..., 0.0049, -0.0371, 0.0017], + [ 0.0075, -0.0126, 0.0049, ..., -0.0204, 0.0041, 0.0056], + [-0.0110, -0.0195, 0.0056, ..., -0.0542, -0.0092, -0.0053], + ..., + [-0.0205, -0.0085, -0.0041, ..., -0.0199, 0.0041, 0.0061], + [-0.0220, -0.0302, -0.0099, ..., 0.0104, 0.0369, 0.0043], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[-0.0023, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0047, 0.0049, 0.0238, ..., -0.0145, -0.0320, 0.0037], + [-0.0193, 0.0364, -0.0048, ..., -0.0064, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0048, -0.0232, -0.0464], + [-0.0109, -0.0298, -0.0195, ..., 0.0114, 0.0312, 0.0156], + [-0.0074, -0.0117, -0.0195, ..., 0.0043, 0.0106, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0048, -0.0239, -0.0123, ..., -0.0045, 0.0243, -0.0396], + [-0.0101, 0.0155, -0.0056, ..., -0.0277, -0.0038, 0.0037], + ..., + [ 0.0117, -0.0057, 0.0515, ..., 0.0327, -0.0120, -0.0066], + [ 0.0041, 0.0170, 0.0074, ..., -0.0048, 0.0280, -0.0210], + [ 0.0036, 0.0068, 0.0057, ..., 0.0153, -0.0077, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0061, 0.0081, -0.0077, ..., -0.0588, 0.0187, -0.0064], + [-0.0102, 0.0199, -0.0077, ..., 0.0080, -0.0332, 0.0040], + ..., + [ 0.0078, -0.0464, 0.0277, ..., -0.0145, 0.0198, -0.0078], + [ 0.0162, 0.0038, 0.0090, ..., -0.0109, 0.0081, 0.0297], + [-0.0269, 0.0135, 0.0164, ..., 0.0079, 0.0176, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0042, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, -0.0026, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, 0.0048, -0.0201, ..., 0.0330, 0.0309, -0.0051], + ..., + [ 0.0039, 0.0123, 0.0153, ..., 0.0039, -0.0039, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0040, 0.0175], + [-0.0244, -0.0059, -0.0300, ..., 0.0049, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0045, -0.0041, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0038, ..., 0.0310, 0.0088, -0.0039], + [ 0.0025, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0110, -0.0237, ..., -0.0104, 0.0157, 0.0078], + [-0.0070, 0.0205, 0.0143, ..., 0.0262, -0.0195, 0.0095], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0078]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0024, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0022, 0.0034, -0.0093], + [ 0.0017, -0.0010, 0.0398, ..., -0.0304, -0.0028, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0039], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0051, 0.0291], + [ 0.0132, 0.0038, -0.0339, ..., 0.0077, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, -0.0027], + [-0.0166, -0.0036, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0061, -0.0032, ..., 0.0240, 0.0036, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0019], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0007, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0067, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0040], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, 0.0040, -0.0039], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, -0.0039, ..., -0.0164, -0.0039, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0026, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0025, 0.0226], + ..., + [ 0.0053, -0.0231, -0.0255, ..., -0.0052, 0.0295, -0.0058], + [-0.0255, 0.0513, 0.0039, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0040, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0056, 0.0078, -0.0364, ..., -0.0175, 0.0306, -0.0075], + [-0.0051, -0.0195, -0.0016, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0070, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0078, -0.0193, -0.0140], + [-0.0166, 0.0059, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0138, 0.0082, 0.0146, ..., -0.0081, 0.0211, -0.0270], + [ 0.0063, -0.0077, -0.0234, ..., 0.0089, -0.0055, -0.0165], + [ 0.0225, -0.0251, 0.0047, ..., 0.0072, 0.0236, -0.0036], + ..., + [ 0.0186, 0.0148, -0.0053, ..., 0.0272, -0.0062, -0.0086], + [ 0.0064, 0.0029, 0.0078, ..., -0.0260, 0.0267, 0.0147], + [ 0.0101, -0.0044, 0.0212, ..., 0.0195, -0.0157, 0.0114]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0084, -0.0270, 0.0251, ..., 0.0098, 0.0055, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.2.proj.1.weight': tensor([[ 0.0036, 0.0231, 0.0918, ..., -0.0571, 0.0250, -0.0284], + [-0.0237, 0.0006, -0.1128, ..., 0.0193, -0.0238, 0.0518], + [-0.0087, 0.1279, -0.0123, ..., -0.0220, 0.0184, -0.0349], + ..., + [ 0.0146, 0.0055, -0.0076, ..., 0.0234, -0.0280, 0.0082], + [-0.0237, -0.0608, 0.0603, ..., 0.0332, 0.0708, 0.0505], + [ 0.0146, -0.0229, -0.0050, ..., 0.0249, 0.0420, -0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_proj.weight': tensor([[-0.0093, 0.0068, 0.0118, ..., -0.0078, -0.0047, 0.0182], + [ 0.0018, 0.0015, -0.0032, ..., -0.0032, -0.0005, -0.0092], + [-0.0064, 0.0067, -0.0002, ..., 0.0002, -0.0104, -0.0227], + ..., + [ 0.0010, -0.0172, -0.0045, ..., 0.0082, 0.0030, -0.0064], + [-0.0014, 0.0045, -0.0045, ..., -0.0121, -0.0164, -0.0048], + [ 0.0080, -0.0050, -0.0016, ..., 0.0063, 0.0061, 0.0010]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_norm.weight': tensor([ 1.4062e-01, -4.2480e-02, 8.6426e-02, 2.3145e-01, -3.4485e-03, + -8.3160e-04, -1.3933e-06, -1.8794e-21, 6.6174e-22, -2.0117e-20, + -9.3408e-33, 3.3352e-21, -6.1512e-05, -1.1942e-23, -2.9565e-25, + 2.0954e-30, 1.8597e-05, 2.6484e-09, -3.7517e-12, -1.8406e-04, + 2.4738e-10, -4.5528e-21, 4.1504e-02, -3.2616e-04, -1.0742e-01, + 5.8594e-01, 5.4932e-02, -7.3853e-03, 4.9805e-02, 9.8633e-02, + -6.3019e-19, -1.4095e-21, 1.0757e-03, -9.5367e-05, 1.2346e-13, + -2.4033e-04, 3.4106e-12, -7.1086e-25, 1.1529e-23, -2.0292e-24, + -2.3588e-25, -6.0988e-26, 1.1228e-25, -6.5308e-03, 4.2969e-02, + -1.4771e-02, 2.9785e-02, 1.2354e-01, 3.4180e-01, 7.8125e-03, + -4.6082e-03, 5.3406e-03, 4.4556e-03, 1.2005e-10, 2.8922e-10, + 1.4095e-17, -4.8307e-22, -1.0503e-18, 3.0494e-26, -2.6496e-24, + 5.1670e-29, -8.8938e-20, 2.1487e-25, 2.4536e-14, 2.7148e-01, + 2.2559e-01, 3.7109e-01, -4.9316e-02, 2.6855e-03, 7.1335e-04, + 1.0282e-06, 4.3675e-22, 1.2811e-20, 5.7175e-20, -1.3482e-31, + -7.3123e-22, 3.6163e-03, -7.3412e-24, -2.1306e-26, 6.2320e-29, + 1.9226e-03, -9.7316e-11, 4.0927e-11, -3.8452e-03, 3.3253e-12, + -6.6746e-19, -3.9978e-03, -1.2360e-03, 5.7812e-01, 1.1719e-01, + 5.0391e-01, 9.2773e-02, -5.8594e-03, 9.0408e-04, 1.6263e-18, + -5.2675e-21, 3.0060e-03, 4.6730e-05, 4.2188e-14, 3.9978e-03, + -5.7116e-10, -1.4217e-24, -8.4011e-25, -2.8290e-22, -4.2914e-28, + -2.5022e-23, 8.5591e-29, 1.4404e-02, 5.4626e-03, -6.0425e-03, + 4.1211e-01, 2.7734e-01, 2.0410e-01, -1.1047e-02, -5.2185e-03, + 7.8125e-03, 4.6692e-03, -4.8885e-11, 1.4461e-10, -9.1480e-19, + 4.1558e-21, 1.2197e-18, 3.3120e-26, -3.9911e-23, -1.2572e-29, + -3.2591e-22, -1.7707e-24, 8.5038e-11], dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.self_attn.k_proj.weight': tensor([[-0.0053, 0.0036, -0.0114, ..., -0.0015, -0.0072, 0.0092], + [-0.0233, -0.0015, 0.0042, ..., -0.0058, -0.0069, -0.0112], + [-0.0036, -0.0111, -0.0016, ..., 0.0052, 0.0021, 0.0089], + ..., + [ 0.0049, -0.0039, -0.0018, ..., -0.0025, 0.0008, 0.0016], + [ 0.0049, 0.0009, 0.0007, ..., -0.0044, 0.0020, 0.0024], + [ 0.0040, 0.0023, 0.0050, ..., -0.0034, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.k_norm.weight': tensor([ 2.4121e-01, 1.2598e-01, 6.2012e-02, 2.0898e-01, 3.3417e-03, + -1.1292e-03, 1.4082e-06, -1.9852e-21, 2.2896e-21, 2.8164e-20, + 1.9548e-32, 3.2955e-21, -9.9659e-05, -1.1581e-23, -2.9404e-25, + 2.7364e-30, -7.3016e-07, 2.6921e-09, 2.0464e-12, 1.9360e-04, + -2.2555e-10, 2.7926e-21, 3.7354e-02, -9.9182e-04, 7.6172e-02, + 5.8594e-01, 2.2852e-01, 8.9844e-02, 6.2256e-03, 6.8970e-03, + 2.5072e-19, 9.3306e-22, -3.7003e-04, -9.8228e-05, 1.4122e-13, + -1.0824e-04, -3.0695e-11, -1.3248e-25, -1.1684e-23, 1.0792e-24, + -2.2941e-25, -1.1551e-25, -1.1228e-25, 1.0681e-02, 4.4678e-02, + -1.1536e-02, 2.2852e-01, 2.0996e-01, 3.7109e-01, 1.2085e-02, + -4.2725e-03, -7.8125e-03, 7.6904e-03, -1.2824e-10, 3.1650e-10, + -7.5352e-18, -1.7999e-21, 1.3281e-18, 4.2005e-26, -2.7271e-24, + -5.4037e-29, -8.6821e-20, -2.1487e-25, 2.6867e-14, -1.0547e-01, + 2.2168e-01, 3.7305e-01, -1.5430e-01, -2.1667e-03, -9.6321e-05, + -1.0431e-06, -6.3693e-23, 1.2705e-20, 5.5904e-20, 1.3173e-31, + 1.0059e-21, -3.5858e-03, 6.2814e-24, -2.0700e-26, 6.6659e-29, + 1.9302e-03, 2.0191e-10, 4.2292e-11, -3.8452e-03, -2.7427e-12, + -6.5391e-19, -7.8735e-03, -2.8229e-04, 5.7812e-01, -1.3477e-01, + 4.9609e-01, 7.0801e-03, 4.9072e-02, 9.4727e-02, -1.8431e-18, + -5.2675e-21, -3.1891e-03, 2.1338e-05, -1.9207e-14, -3.9978e-03, + 5.8208e-10, -1.4282e-24, 3.2829e-24, -2.7132e-22, 1.6787e-27, + -1.5303e-23, 2.9819e-28, 9.4604e-03, 2.0294e-03, 1.0559e-02, + 4.0430e-01, 2.3340e-01, 1.8555e-01, 8.2397e-03, 5.9814e-03, + 4.6387e-03, -1.2054e-03, 1.6257e-11, 2.6489e-11, -8.4026e-18, + 3.0308e-21, -2.2192e-19, -2.1104e-26, 3.8877e-23, -1.0748e-29, + 6.0880e-21, -1.8482e-24, -8.4583e-11], dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.self_attn.v_proj.weight': tensor([[-0.0007, 0.0037, -0.0014, ..., 0.0045, 0.0013, -0.0028], + [ 0.0083, 0.0014, -0.0035, ..., -0.0039, 0.0026, 0.0009], + [ 0.0136, 0.0058, 0.0026, ..., -0.0038, -0.0015, 0.0028], + ..., + [-0.0036, -0.0049, 0.0043, ..., 0.0056, -0.0034, -0.0077], + [-0.0043, -0.0018, -0.0001, ..., 0.0055, -0.0032, -0.0046], + [-0.0015, 0.0055, -0.0029, ..., 0.0037, 0.0025, -0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.output_proj.weight': tensor([[ 8.4229e-03, -3.6133e-02, -1.4099e-02, ..., -6.5918e-03, + 5.1270e-03, 5.0659e-03], + [ 6.6223e-03, -7.2327e-03, 5.8594e-03, ..., -2.6703e-03, + 1.8311e-03, 3.2043e-03], + [-3.2196e-03, 5.2214e-05, 7.5684e-03, ..., 4.5776e-03, + 4.5166e-03, 2.8839e-03], + ..., + [ 2.7008e-03, 5.2490e-03, 7.0801e-03, ..., -4.4250e-03, + -6.4697e-03, -6.5002e-03], + [-2.2125e-03, 5.1270e-03, -6.3477e-03, ..., -9.4604e-03, + 8.1787e-03, 4.3640e-03], + [-4.2114e-03, 1.0132e-02, 3.9368e-03, ..., 4.6997e-03, + 4.4861e-03, 4.4556e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_proj.weight': tensor([[ 0.0085, 0.0082, -0.0034, ..., -0.0069, 0.0062, -0.0014], + [ 0.0073, -0.0261, -0.0047, ..., 0.0066, -0.0082, 0.0026], + [-0.0121, 0.0070, -0.0002, ..., 0.0057, 0.0209, -0.0048], + ..., + [ 0.0033, 0.0048, -0.0010, ..., -0.0012, -0.0046, 0.0108], + [-0.0066, -0.0033, 0.0006, ..., 0.0024, 0.0063, -0.0033], + [-0.0024, -0.0065, 0.0020, ..., -0.0033, -0.0007, 0.0029]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_norm.weight': tensor([ 9.7275e-04, 1.0315e-02, 3.7107e-10, 2.0752e-03, 2.0294e-03, + 2.3723e-05, -1.9550e-05, 5.0964e-03, 3.4180e-02, 2.6001e-02, + 6.8918e-07, 2.4536e-02, 5.5469e-01, 7.4768e-03, 8.4229e-03, + 5.3711e-03, 4.6875e-01, 1.6689e-05, 3.6163e-03, 1.1265e-05, + 5.1498e-04, 4.2725e-03, 5.9814e-02, 1.5991e-02, 1.1963e-02, + 6.3171e-03, 2.5146e-02, -7.5817e-05, 9.9487e-03, 6.7444e-03, + 4.1602e-01, 2.5146e-02, 4.8256e-04, 4.2188e-01, 5.3125e-01, + 1.9684e-03, 3.4790e-03, 3.4668e-02, 1.1475e-01, 1.4893e-02, + 3.8818e-02, 4.0894e-03, 6.4028e-09, 1.4496e-03, 1.8921e-02, + 2.6855e-02, 1.1719e-02, 9.7656e-03, 1.4954e-03, 9.8348e-06, + 8.5449e-03, -9.7156e-06, -7.1526e-06, 1.6809e-05, -4.6539e-04, + -2.6941e-05, 1.8677e-02, 1.0620e-02, 1.0620e-02, 2.9907e-03, + 4.0054e-04, 1.8716e-05, 8.7280e-03, -1.3530e-05, 1.3828e-04, + 1.4782e-04, 1.7047e-05, 2.5391e-01, 7.4387e-04, 1.7090e-02, + 4.0430e-01, -1.9455e-03, 3.9307e-02, 4.9744e-03, 5.4240e-06, + 1.3351e-04, 9.6436e-03, 4.9591e-04, 2.9449e-03, 1.8188e-02, + 2.0386e-02, 1.6968e-02, 1.9043e-01, 6.1951e-03, 3.4094e-05, + 1.5564e-02, 5.4321e-03, 2.0862e-05, 6.1512e-05, 4.7656e-01, + 7.6599e-03, -8.3447e-05, 1.0193e-02, 3.0100e-06, -5.2691e-05, + 4.3555e-01, 2.8419e-04, 7.9346e-03, -1.8304e-11, 1.0834e-03, + 4.1797e-01, 1.7166e-05, 9.5825e-03, -4.8578e-06, 8.3618e-03, + -3.4571e-05, 2.0409e-04, -2.4033e-04, 2.3804e-02, 1.1597e-02, + 6.1951e-03, -4.7982e-06, 3.1494e-02, 1.1301e-04, 1.1841e-02, + 1.4114e-04, 2.5513e-02, 4.1504e-03, 6.7749e-03, 4.1199e-04, + 2.5272e-05, 1.1169e-02, 8.7280e-03, 6.9885e-03, 1.5198e-02, + 2.6512e-04, 2.6855e-03, 5.3167e-05], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_proj.weight': tensor([[-0.0008, 0.0007, 0.0026, ..., -0.0004, -0.0014, 0.0005], + [-0.0063, -0.0021, -0.0051, ..., -0.0028, -0.0053, -0.0034], + [-0.0005, -0.0015, -0.0016, ..., -0.0027, -0.0029, -0.0014], + ..., + [ 0.0026, -0.0025, 0.0018, ..., 0.0011, 0.0024, -0.0006], + [ 0.0010, 0.0005, -0.0018, ..., -0.0025, 0.0009, 0.0019], + [ 0.0035, 0.0019, -0.0023, ..., -0.0041, -0.0006, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_norm.weight': tensor([-9.7275e-04, 1.0254e-02, 3.6016e-10, 1.8921e-03, 2.0142e-03, + 2.7061e-05, 1.9550e-05, 5.0659e-03, 3.3203e-02, 2.6123e-02, + -7.1898e-07, 2.4902e-02, 5.5859e-01, 7.4463e-03, 7.9346e-03, + 5.3406e-03, 4.6875e-01, 1.7524e-05, 3.5248e-03, -2.5392e-05, + 5.7983e-04, 4.3030e-03, 5.9814e-02, 1.5991e-02, 1.1780e-02, + 6.3782e-03, 2.5024e-02, -7.5817e-05, 9.3384e-03, 6.7444e-03, + 4.1602e-01, 2.5024e-02, 4.5395e-04, 4.2188e-01, 5.3125e-01, + 1.8234e-03, 3.5553e-03, 3.4912e-02, 1.1572e-01, 1.4893e-02, + 3.8818e-02, 4.1199e-03, -6.2282e-09, 1.2131e-03, 1.8555e-02, + 2.6733e-02, 1.1719e-02, 9.7656e-03, 1.3580e-03, 9.8944e-06, + 8.4839e-03, -1.0431e-05, -7.1526e-06, 1.9789e-05, 4.6539e-04, + 1.3947e-05, 1.8433e-02, 1.0742e-02, 1.0925e-02, 2.9907e-03, + -2.8419e-04, 1.8716e-05, 8.4229e-03, -1.3888e-05, 8.6308e-05, + 1.8692e-04, 1.6928e-05, 2.5391e-01, 6.4468e-04, 1.7090e-02, + 4.0430e-01, -1.9455e-03, 3.7842e-02, 4.9438e-03, -1.7956e-06, + 1.6117e-04, 8.6670e-03, 4.7112e-04, 2.9297e-03, 1.8188e-02, + 2.0264e-02, 1.7090e-02, 1.9043e-01, 6.0425e-03, 2.6345e-05, + 1.5564e-02, 5.8289e-03, 2.2173e-05, 5.7459e-05, 4.7656e-01, + 7.6599e-03, -9.9182e-05, 9.8877e-03, 2.2948e-06, -5.4359e-05, + 4.3555e-01, -1.0967e-04, 7.9956e-03, 1.8304e-11, 1.0605e-03, + 4.1797e-01, 1.6928e-05, 1.0498e-02, -7.1526e-06, 8.3618e-03, + -3.5048e-05, 2.2697e-04, -2.2984e-04, 2.3438e-02, 1.1536e-02, + 6.2866e-03, -4.2617e-06, 3.1128e-02, 1.1396e-04, 1.2024e-02, + -1.0920e-04, 2.5513e-02, 4.2725e-03, 7.2021e-03, 4.2534e-04, + 2.5153e-05, 1.1292e-02, 8.7280e-03, 6.9885e-03, 1.4771e-02, + 1.6975e-04, 2.7466e-03, -4.8637e-05], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.v_proj.weight': tensor([[-0.0104, 0.0016, 0.0097, ..., 0.0032, 0.0028, -0.0110], + [-0.0243, -0.0013, 0.0063, ..., -0.0010, -0.0076, 0.0008], + [-0.0132, 0.0032, 0.0044, ..., 0.0031, 0.0153, -0.0128], + ..., + [ 0.0117, 0.0008, 0.0018, ..., -0.0023, 0.0064, 0.0181], + [-0.0046, -0.0002, 0.0117, ..., -0.0018, -0.0161, -0.0034], + [-0.0216, -0.0033, 0.0044, ..., -0.0035, -0.0064, -0.0221]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.output_proj.weight': tensor([[ 0.0051, -0.0060, -0.0038, ..., -0.0053, -0.0006, 0.0098], + [ 0.0043, -0.0039, -0.0088, ..., 0.0214, -0.0043, -0.0114], + [-0.0044, 0.0049, 0.0037, ..., 0.0083, 0.0037, 0.0073], + ..., + [-0.0089, -0.0048, 0.0079, ..., -0.0076, -0.0035, -0.0005], + [ 0.0046, -0.0064, 0.0063, ..., 0.0055, -0.0054, -0.0113], + [ 0.0150, -0.0101, -0.0053, ..., 0.0051, -0.0061, 0.0074]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_img.weight': tensor([[ 0.0053, 0.0117, 0.0039, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0072, 0.0249, 0.0291], + [-0.0095, -0.0042, -0.0115, ..., -0.0342, -0.0327, 0.0066], + ..., + [ 0.0078, 0.0045, 0.0045, ..., 0.0086, 0.0286, -0.0165], + [ 0.0068, -0.0053, -0.0101, ..., -0.0101, 0.0142, -0.0309], + [ 0.0242, -0.0102, 0.0255, ..., -0.0227, -0.0044, -0.0153]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, 0.0067], + [ 0.0042, 0.0121, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0084, ..., -0.0059, -0.0057, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0037, 0.0049], + [-0.0160, 0.0051, 0.0039, ..., 0.0133, -0.0173, -0.0046], + [-0.0051, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0098]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0048, -0.0364, ..., -0.0225, -0.0142, -0.0056], + [-0.0157, 0.0051, -0.0364, ..., 0.0101, 0.0120, 0.0036], + [ 0.0043, 0.0615, 0.0010, ..., -0.0205, -0.0306, -0.0115], + ..., + [-0.0038, -0.0038, -0.0103, ..., -0.0211, -0.0019, 0.0288], + [ 0.0266, -0.0123, -0.0044, ..., -0.0420, -0.0282, 0.0053], + [ 0.0286, 0.0325, 0.0093, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.mlp.layer1.weight': tensor([[ 0.0073, 0.0172, -0.0126, ..., 0.0037, -0.0081, 0.0063], + [-0.0317, 0.0056, -0.0040, ..., -0.0054, 0.0044, -0.0088], + [-0.0048, -0.0020, 0.0048, ..., 0.0035, 0.0062, -0.0051], + ..., + [-0.0073, -0.0065, -0.0038, ..., -0.0095, 0.0120, 0.0034], + [ 0.0161, 0.0081, 0.0046, ..., -0.0063, 0.0014, 0.0034], + [-0.0178, -0.0041, -0.0038, ..., 0.0040, -0.0090, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.mlp.layer2.weight': tensor([[ 0.0001, -0.0159, 0.0052, ..., -0.0053, 0.0208, -0.0045], + [-0.0054, -0.0058, -0.0045, ..., -0.0073, 0.0044, -0.0053], + [ 0.0075, 0.0032, 0.0086, ..., 0.0044, 0.0024, -0.0157], + ..., + [ 0.0098, -0.0069, -0.0067, ..., 0.0011, 0.0050, -0.0056], + [-0.0063, 0.0007, 0.0046, ..., 0.0077, 0.0239, -0.0121], + [ 0.0059, -0.0151, 0.0050, ..., 0.0119, -0.0114, -0.0085]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0576, -0.0562, -0.0547, ..., 0.0029, 0.0030, 0.0030], + [ 0.0559, 0.0549, 0.0532, ..., -0.0023, -0.0024, -0.0025], + [ 0.0554, 0.0544, 0.0532, ..., -0.0025, -0.0025, -0.0026], + ..., + [ 0.0566, 0.0557, 0.0535, ..., -0.0022, -0.0024, -0.0024], + [ 0.0535, 0.0525, 0.0505, ..., -0.0021, -0.0021, -0.0021], + [ 0.0571, 0.0559, 0.0547, ..., -0.0029, -0.0030, -0.0029]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_self_attn.2.weight': tensor([[ 0.0019, -0.0016, -0.0020, ..., -0.0028, -0.0024, -0.0018], + [ 0.0017, -0.0021, -0.0018, ..., 0.0024, -0.0017, -0.0018], + [ 0.0002, 0.0033, 0.0002, ..., 0.0035, 0.0034, -0.0003], + ..., + [-0.0374, 0.0452, 0.0442, ..., 0.0420, 0.0374, 0.0374], + [-0.0100, 0.0103, 0.0104, ..., 0.0090, 0.0061, 0.0103], + [ 0.0078, -0.0095, -0.0092, ..., -0.0069, -0.0066, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_cross_attn.1.weight': tensor([[-9.0942e-03, -8.6670e-03, 7.8125e-03, ..., -6.0654e-04, + -5.3406e-04, -7.0572e-04], + [-4.1992e-02, -4.0527e-02, -4.0039e-02, ..., 2.9945e-04, + 3.0518e-04, 3.1471e-04], + [ 4.3213e-02, 4.2236e-02, 4.1748e-02, ..., -8.7738e-05, + -6.7711e-05, -9.2506e-05], + ..., + [ 4.1748e-02, 4.0771e-02, 4.0283e-02, ..., -3.7766e-04, + -3.1090e-04, -3.4714e-04], + [ 5.4688e-02, 5.3223e-02, 5.2002e-02, ..., -2.0981e-04, + -1.3924e-04, -1.9932e-04], + [ 3.8574e-02, 3.6865e-02, 3.6377e-02, ..., -4.7386e-06, + 8.0585e-05, 3.0994e-05]], dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0184, -0.0074, 0.0074, ..., 0.0074, 0.0109, 0.0067], + [-0.0044, -0.0042, 0.0042, ..., 0.0042, 0.0062, 0.0042], + [-0.0007, 0.0049, -0.0077, ..., -0.0049, -0.0074, -0.0078], + ..., + [-0.0078, -0.0265, 0.0332, ..., 0.0265, 0.0432, 0.0286], + [ 0.0114, 0.0063, 0.0122, ..., -0.0059, 0.0040, 0.0134], + [-0.0136, 0.0040, -0.0077, ..., -0.0040, -0.0046, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_mlp.1.weight': tensor([[-0.0037, -0.0033, -0.0031, ..., 0.0002, 0.0004, 0.0004], + [ 0.0540, 0.0518, 0.0503, ..., -0.0020, -0.0023, -0.0024], + [ 0.0459, 0.0444, 0.0437, ..., -0.0033, -0.0036, -0.0036], + ..., + [ 0.0884, 0.0854, 0.0845, ..., -0.0005, -0.0007, -0.0008], + [ 0.0035, 0.0035, 0.0036, ..., -0.0002, -0.0003, -0.0002], + [ 0.0654, 0.0635, 0.0618, ..., -0.0010, -0.0010, -0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_mlp.2.weight': tensor([[-0.0004, 0.0125, 0.0136, ..., 0.0120, 0.0003, 0.0098], + [-0.0039, -0.0009, 0.0010, ..., -0.0024, 0.0039, -0.0014], + [-0.0015, 0.0054, 0.0057, ..., 0.0036, 0.0015, 0.0044], + ..., + [ 0.0084, 0.0654, 0.0845, ..., 0.0688, -0.0084, 0.0500], + [ 0.0065, -0.0017, -0.0024, ..., -0.0022, -0.0064, -0.0020], + [-0.0120, -0.0071, -0.0072, ..., -0.0064, 0.0119, -0.0076]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.before_proj.weight': tensor([[ 0.0081, 0.0043, -0.0024, ..., -0.0027, -0.0028, 0.0038], + [-0.0078, -0.0134, -0.0053, ..., 0.0205, 0.0040, 0.0084], + [-0.0042, -0.0051, 0.0022, ..., 0.0023, 0.0043, -0.0065], + ..., + [-0.0031, 0.0089, 0.0085, ..., 0.0024, -0.0022, -0.0054], + [ 0.0143, 0.0198, 0.0146, ..., 0.0020, 0.0107, 0.0045], + [-0.0003, -0.0062, 0.0044, ..., 0.0002, -0.0022, -0.0056]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.before_proj.bias': tensor([ 0.0018, 0.0023, 0.0036, ..., 0.0046, 0.0033, -0.0115], + dtype=torch.bfloat16), 'net.control_blocks_2.0.after_proj.weight': tensor([[-0.0176, 0.0060, -0.0047, ..., -0.0038, -0.0049, -0.0137], + [-0.0106, -0.0037, -0.0068, ..., -0.0107, 0.0072, 0.0188], + [ 0.0019, 0.0009, -0.0058, ..., -0.0017, 0.0006, -0.0014], + ..., + [-0.0062, -0.0015, -0.0006, ..., 0.0024, 0.0011, -0.0017], + [ 0.0145, -0.0029, 0.0059, ..., -0.0047, -0.0109, 0.0077], + [ 0.0077, 0.0045, -0.0050, ..., 0.0039, -0.0055, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.after_proj.bias': tensor([ 0.0300, -0.0136, 0.0002, ..., -0.0012, -0.0262, 0.0337], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_proj.weight': tensor([[-0.0199, 0.0188, 0.0041, ..., 0.0259, -0.0085, 0.0050], + [-0.0106, -0.0162, 0.0114, ..., -0.0073, -0.0008, -0.0393], + [ 0.0139, 0.0078, 0.0051, ..., 0.0129, 0.0022, -0.0157], + ..., + [ 0.0159, 0.0104, -0.0030, ..., 0.0101, -0.0117, -0.0049], + [-0.0114, -0.0149, -0.0050, ..., -0.0016, -0.0167, -0.0042], + [-0.0027, -0.0081, -0.0071, ..., -0.0084, 0.0095, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_norm.weight': tensor([ 3.0664e-01, 9.6875e-01, 5.7129e-02, 7.6562e-01, 9.4922e-01, + -6.4319e-09, -3.0640e-02, 5.4123e-15, -4.0149e-19, 5.0093e-13, + 3.9916e-28, -8.3778e-33, -5.4919e-35, 3.3253e-22, -2.3074e-29, + 3.4907e-34, 1.4941e-01, 2.9325e-05, 1.5068e-04, -1.8921e-02, + 5.1880e-03, -1.3447e-04, 2.6367e-01, 9.4922e-01, 1.1875e+00, + 3.4766e-01, 7.3547e-03, 4.2383e-01, 1.7090e-02, 4.6327e-12, + 1.3281e-17, 4.6629e-15, 5.6839e-04, 1.4343e-03, -3.5095e-04, + 1.3184e-02, 1.4782e-04, 2.6375e-10, -3.8338e-04, 4.2848e-16, + 2.6455e-17, -2.0539e-14, -3.1896e-21, 8.8281e-01, 2.7148e-01, + 2.0117e-01, 4.7852e-01, 1.1406e+00, 1.2402e-01, -3.6926e-03, + -2.1118e-02, 2.0020e-01, 2.8992e-04, 1.8917e-10, -3.8147e-06, + 7.4593e-16, 7.2632e-03, 4.1034e-13, 7.5731e-29, -6.0486e-34, + 3.8556e-35, 1.6515e-15, -1.2494e-20, 1.0532e-33, 3.4766e-01, + 2.5586e-01, 5.8203e-01, 2.0996e-01, 1.1914e-01, -3.4925e-09, + 1.9287e-02, 2.2898e-15, -1.7364e-19, -5.2935e-13, -7.8097e-29, + 8.7269e-34, 2.7836e-35, -9.9262e-23, -1.1107e-27, -1.1059e-34, + 1.4191e-03, 2.7466e-04, 1.6403e-04, -1.2970e-04, 3.0708e-04, + 4.2236e-02, 9.4922e-01, 2.1582e-01, 2.0312e-01, 1.0781e+00, + 1.9824e-01, 1.8945e-01, -4.1504e-03, -4.0075e-12, -3.6592e-18, + 3.8303e-15, 1.0757e-03, 1.6724e-02, -3.8910e-03, -4.3640e-03, + 3.3188e-04, 5.5252e-11, 2.2799e-06, -7.3726e-18, 6.6058e-15, + -5.0848e-14, -5.5318e-24, 2.0312e-01, 1.1484e+00, 1.1953e+00, + 1.3750e+00, 4.0234e-01, 3.8281e-01, 9.8877e-03, -2.1484e-02, + -2.2583e-02, -7.4863e-05, 1.9554e-10, 2.4915e-05, 1.3791e-16, + 1.6846e-02, 1.0747e-13, 2.2404e-28, -6.4398e-34, -1.0391e-35, + -1.3214e-19, 7.1998e-20, 7.7037e-34], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.self_attn.k_proj.weight': tensor([[ 0.0152, 0.0017, -0.0059, ..., -0.0053, 0.0058, -0.0119], + [ 0.0231, 0.0071, 0.0044, ..., -0.0135, -0.0038, -0.0149], + [ 0.0060, 0.0164, -0.0024, ..., -0.0013, 0.0027, 0.0041], + ..., + [-0.0061, 0.0035, -0.0077, ..., 0.0115, 0.0005, -0.0031], + [-0.0100, -0.0090, -0.0032, ..., 0.0115, -0.0021, -0.0073], + [-0.0007, -0.0018, -0.0008, ..., -0.0059, 0.0046, 0.0048]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.k_norm.weight': tensor([ 3.2617e-01, 9.8047e-01, -1.0205e-01, 7.8516e-01, 9.4922e-01, + 2.9686e-09, 2.6245e-02, -5.9119e-15, -3.9980e-19, 5.4357e-13, + 4.5438e-28, -3.2259e-33, 4.4387e-35, 3.3418e-22, -2.1201e-29, + 3.4155e-34, 1.4941e-01, 2.9206e-05, -7.9155e-05, 5.4932e-02, + 5.1880e-03, 8.8215e-05, 2.9492e-01, 9.5312e-01, 1.1719e+00, + 1.4258e-01, 1.4844e-01, 4.6289e-01, 1.4404e-02, -4.5759e-12, + 1.3390e-17, 4.4964e-15, 5.3406e-04, -1.9302e-03, -3.2806e-04, + 4.2236e-02, 1.4687e-04, 2.1009e-10, -2.3651e-04, -4.2848e-16, + 2.4395e-17, -1.9540e-14, -3.1896e-21, 8.7891e-01, 2.5586e-01, + -4.7302e-03, -3.2227e-01, 1.1484e+00, 1.3672e-01, -3.0518e-03, + 2.8809e-02, 1.2402e-01, 1.0633e-04, -1.5280e-10, -3.4720e-06, + -7.4593e-16, 8.4229e-03, 4.1034e-13, 6.9814e-29, 9.3287e-35, + -1.2413e-35, -1.6515e-15, -1.1911e-20, -1.9410e-34, 3.8281e-01, + 3.1250e-01, 5.8594e-01, 1.6895e-01, -9.9609e-02, -4.1036e-09, + -2.2827e-02, -1.5543e-15, 4.5401e-19, -5.7909e-13, 2.1299e-29, + 1.0171e-33, 1.9184e-35, 8.2718e-23, 1.0539e-27, -4.8750e-34, + 3.4332e-04, 3.2663e-05, -1.6594e-04, -1.0920e-04, 3.0899e-04, + 4.2236e-02, 9.4531e-01, 2.0898e-01, 1.9824e-01, 1.0703e+00, + 1.8555e-01, 2.2266e-01, 9.3994e-03, 3.8085e-12, -4.4181e-18, + -3.6915e-15, 1.2054e-03, -1.5747e-02, -3.8910e-03, 3.7598e-02, + 3.2997e-04, 2.8922e-10, 1.6317e-06, -4.7976e-18, 6.6891e-15, + -5.0848e-14, 5.8678e-24, 2.9297e-01, 1.1406e+00, 1.1953e+00, + 1.3750e+00, 1.2891e-01, 3.9648e-01, -9.7656e-03, 4.3030e-03, + -1.6992e-01, 3.4332e-04, -1.7280e-10, -2.5034e-05, -1.3618e-16, + -1.5991e-02, -1.0569e-13, 1.9958e-28, -4.8750e-34, -3.7616e-35, + 3.3475e-18, 7.1574e-20, -4.5139e-34], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.self_attn.v_proj.weight': tensor([[ 8.2397e-03, 5.5847e-03, 7.1106e-03, ..., -2.0630e-02, + 8.9722e-03, 4.8828e-03], + [ 1.1108e-02, 6.4697e-03, 8.4686e-04, ..., -3.6163e-03, + -4.8523e-03, -4.3335e-03], + [-2.0905e-03, -3.7079e-03, 9.7275e-05, ..., -6.8359e-03, + 4.9744e-03, -7.9346e-03], + ..., + [ 3.6926e-03, 8.4229e-03, 1.3428e-02, ..., 2.0508e-02, + 3.6926e-03, -1.7578e-02], + [-1.8555e-02, 5.4321e-03, 1.1169e-02, ..., 3.5889e-02, + -7.6294e-03, 6.4850e-04], + [-5.1880e-03, -4.0588e-03, -5.6152e-03, ..., 1.9531e-02, + -5.6458e-03, 4.4861e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.output_proj.weight': tensor([[-0.0062, -0.0013, -0.0136, ..., 0.0078, -0.0044, -0.0070], + [ 0.0066, 0.0037, 0.0198, ..., 0.0023, -0.0028, 0.0106], + [-0.0044, -0.0047, -0.0019, ..., -0.0060, 0.0150, 0.0024], + ..., + [ 0.0079, 0.0010, -0.0066, ..., -0.0031, -0.0305, -0.0067], + [ 0.0038, -0.0014, -0.0115, ..., -0.0028, -0.0035, -0.0024], + [-0.0033, 0.0103, -0.0242, ..., 0.0129, 0.0017, -0.0036]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_proj.weight': tensor([[ 0.0050, 0.0261, 0.0086, ..., 0.0053, 0.0153, -0.0013], + [ 0.0194, 0.0076, 0.0065, ..., 0.0080, -0.0079, -0.0320], + [ 0.0176, 0.0128, 0.0025, ..., -0.0038, -0.0118, -0.0253], + ..., + [ 0.0061, 0.0032, -0.0038, ..., 0.0105, -0.0020, 0.0168], + [ 0.0058, 0.0090, 0.0104, ..., -0.0047, -0.0013, 0.0248], + [ 0.0175, -0.0170, -0.0084, ..., 0.0095, 0.0043, -0.0243]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_norm.weight': tensor([ 1.8799e-02, 4.7070e-01, 2.3560e-02, 1.0315e-02, 1.2107e-07, + 6.2937e-10, 3.8338e-04, 5.0391e-01, 8.6212e-04, 4.1602e-01, + 1.9165e-02, 1.9287e-02, 3.5095e-03, 1.3000e-02, 2.3315e-02, + 1.3977e-02, 2.6733e-02, -3.8147e-05, -4.0293e-05, 3.5400e-02, + 2.6367e-01, 3.2227e-02, 1.7700e-03, 1.4343e-02, 3.5400e-03, + 4.1008e-05, 1.8716e-05, 4.5410e-02, -6.2883e-06, 3.8281e-01, + 8.7738e-05, -1.3828e-04, 6.3171e-03, -2.5034e-05, 1.0864e-02, + 1.0193e-02, -2.0646e-10, 4.1406e-01, 1.0803e-02, 2.5635e-03, + 7.7148e-02, 2.8534e-03, 9.8877e-03, 1.0490e-04, 7.6904e-03, + 1.3977e-02, 2.9144e-03, 1.7212e-02, 1.4587e-02, 1.4709e-02, + 7.6904e-03, 4.0039e-01, 1.3611e-02, 1.1719e-02, 3.1494e-02, + 4.5538e-05, 9.1309e-02, 1.4709e-02, 1.5564e-02, 1.3672e-02, + 5.4836e-06, 7.6904e-03, 1.1597e-02, -1.4019e-04, 1.1414e-02, + 5.0659e-03, -8.0585e-05, 3.7689e-03, 1.8921e-02, 8.1543e-02, + 5.9128e-04, 6.1646e-03, 5.2979e-02, 4.1748e-02, 2.3956e-03, + 1.2695e-02, 4.0531e-06, 1.8677e-02, 3.1006e-02, 2.0996e-02, + 5.1880e-03, 1.6113e-02, 2.8320e-02, 3.6955e-06, 2.5391e-02, + 9.3994e-03, 3.5400e-03, 3.3203e-02, 6.9275e-03, 2.9602e-03, + 1.6724e-02, 2.0981e-04, 4.2419e-03, 2.1362e-02, 1.6724e-02, + -1.3351e-04, 9.9487e-03, 3.2227e-02, 1.5320e-02, 3.8086e-02, + 3.0762e-02, 2.4780e-02, 1.1169e-02, 6.7711e-05, -4.9360e-08, + 6.4392e-03, 1.8692e-04, 2.0264e-02, 8.1787e-03, 1.2573e-02, + 4.1809e-03, 1.1414e-02, 2.0508e-02, 5.1172e-01, 2.0447e-03, + 9.6436e-03, 1.0376e-02, 5.7068e-03, 7.8735e-03, 7.9590e-02, + 9.2983e-06, 4.5538e-05, 1.6451e-05, 1.2695e-02, 1.5259e-02, + 1.6357e-02, 2.4902e-02, 1.2573e-02], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_proj.weight': tensor([[ 0.0014, -0.0033, -0.0022, ..., -0.0038, 0.0019, -0.0005], + [-0.0085, -0.0035, -0.0021, ..., 0.0098, 0.0134, 0.0070], + [ 0.0055, 0.0036, 0.0045, ..., 0.0072, 0.0049, -0.0034], + ..., + [-0.0057, -0.0049, 0.0025, ..., -0.0029, -0.0009, -0.0032], + [-0.0078, -0.0053, -0.0126, ..., -0.0014, 0.0068, 0.0006], + [-0.0042, -0.0004, -0.0013, ..., 0.0044, -0.0016, -0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_norm.weight': tensor([ 1.9531e-02, 4.7070e-01, 2.3560e-02, 1.0437e-02, 1.2107e-07, + -2.2828e-10, 4.1962e-04, 5.0391e-01, 8.6594e-04, 4.1602e-01, + 2.0020e-02, 1.9287e-02, 3.5858e-03, 1.3000e-02, 2.3193e-02, + 1.3977e-02, 2.7100e-02, -3.8385e-05, -3.9816e-05, 3.5645e-02, + 2.6367e-01, 3.1982e-02, 1.7776e-03, 1.4587e-02, 3.6163e-03, + 4.0054e-05, 1.8597e-05, 4.5410e-02, -3.4124e-06, 3.8281e-01, + 8.7261e-05, -1.3924e-04, 6.1646e-03, -2.4796e-05, 1.0803e-02, + 1.0315e-02, -1.9008e-10, 4.1406e-01, 1.0864e-02, 2.4567e-03, + 7.6172e-02, 2.8839e-03, 9.6436e-03, 1.0633e-04, 7.5989e-03, + 1.3916e-02, 2.8687e-03, 1.7090e-02, 1.4404e-02, 1.5015e-02, + 7.5989e-03, 4.0039e-01, 1.3611e-02, 1.1230e-02, 3.1250e-02, + 4.4346e-05, 9.1309e-02, 1.4771e-02, 1.5259e-02, 1.3794e-02, + 6.3479e-06, 8.0566e-03, 1.1597e-02, -1.4019e-04, 1.1475e-02, + 4.9744e-03, -8.0585e-05, 3.7689e-03, 1.9043e-02, 8.1543e-02, + 5.6839e-04, 6.1340e-03, 5.2979e-02, 4.1748e-02, 2.4261e-03, + 1.2878e-02, 4.9472e-06, 1.8555e-02, 3.0884e-02, 2.1118e-02, + 5.1880e-03, 1.6113e-02, 2.7710e-02, 4.0531e-06, 2.5513e-02, + 9.3994e-03, 3.3875e-03, 3.2959e-02, 7.2632e-03, 3.0365e-03, + 1.6968e-02, 2.1172e-04, 4.3335e-03, 2.1240e-02, 1.6602e-02, + -1.3351e-04, 9.2773e-03, 3.1982e-02, 1.5259e-02, 3.7598e-02, + 3.1128e-02, 2.4780e-02, 1.1230e-02, 6.6280e-05, -1.7229e-08, + 6.4392e-03, 1.9360e-04, 1.9775e-02, 8.3008e-03, 1.2451e-02, + 4.2114e-03, 1.1169e-02, 2.0020e-02, 5.1172e-01, 2.0752e-03, + 9.5215e-03, 1.0315e-02, 5.5847e-03, 7.9956e-03, 7.9590e-02, + 1.3411e-05, 4.4584e-05, 1.6093e-05, 1.3000e-02, 1.5320e-02, + 1.6602e-02, 2.5024e-02, 1.2329e-02], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.v_proj.weight': tensor([[-0.0210, 0.0036, 0.0016, ..., 0.0012, -0.0206, -0.0220], + [-0.0035, 0.0081, 0.0038, ..., 0.0037, 0.0134, 0.0095], + [-0.0422, -0.0024, -0.0076, ..., -0.0079, -0.0286, -0.0074], + ..., + [-0.0376, 0.0016, -0.0150, ..., 0.0014, 0.0220, 0.0054], + [ 0.0082, 0.0015, -0.0062, ..., 0.0018, 0.0128, 0.0084], + [ 0.0112, 0.0050, -0.0072, ..., 0.0006, 0.0069, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.output_proj.weight': tensor([[ 7.2327e-03, 9.3384e-03, 4.6692e-03, ..., -6.8359e-03, + 2.5330e-03, 6.9885e-03], + [-3.9368e-03, -1.3184e-02, 6.2866e-03, ..., -8.1787e-03, + 6.6833e-03, 8.4229e-03], + [ 2.6245e-03, -2.0599e-03, -1.9684e-03, ..., -1.6785e-04, + -6.1417e-04, -7.0190e-03], + ..., + [-6.5918e-03, 6.3896e-05, 4.1809e-03, ..., 5.4626e-03, + 2.7466e-03, 7.8735e-03], + [ 7.6294e-03, 1.1658e-02, -5.0735e-04, ..., -2.6123e-02, + 7.1106e-03, 2.1606e-02], + [ 4.5166e-03, 4.0588e-03, -1.7456e-02, ..., -1.0742e-02, + 6.0120e-03, 7.8735e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_img.weight': tensor([[ 0.0079, 0.0034, 0.0430, ..., -0.0327, 0.0042, 0.0146], + [-0.0278, 0.0280, 0.0249, ..., -0.0046, -0.0378, 0.0271], + [ 0.0142, -0.0120, 0.0312, ..., -0.0074, 0.0253, -0.0118], + ..., + [ 0.0245, -0.0152, 0.0027, ..., -0.0452, 0.0216, -0.0209], + [-0.0071, 0.0074, -0.0471, ..., -0.0101, 0.0157, -0.0083], + [ 0.0085, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0028, -0.0122, -0.0115], + [ 0.0137, 0.0062, -0.0188, ..., 0.0157, 0.0082, -0.0047], + [-0.0306, -0.0051, -0.0273, ..., 0.0103, 0.0337, -0.0048], + ..., + [-0.0071, 0.0003, 0.0063, ..., 0.0209, -0.0106, 0.0081], + [ 0.0194, 0.0210, 0.0123, ..., -0.0146, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0109, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img.weight': tensor([[-0.0003, -0.0076, 0.0173, ..., 0.0084, 0.0356, 0.0217], + [-0.0498, -0.0049, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0104, -0.0013, -0.0139, ..., 0.0041, 0.0269, 0.0082], + ..., + [-0.0045, 0.0120, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0077, -0.0162, ..., -0.0059, -0.0076, -0.0121], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0134, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.mlp.layer1.weight': tensor([[-0.0031, -0.0219, -0.0079, ..., 0.0020, -0.0061, -0.0039], + [ 0.0006, 0.0044, 0.0055, ..., 0.0060, -0.0048, -0.0092], + [-0.0221, 0.0100, 0.0028, ..., -0.0045, -0.0090, -0.0073], + ..., + [ 0.0286, 0.0098, 0.0014, ..., 0.0140, 0.0062, 0.0143], + [-0.0159, -0.0058, 0.0041, ..., 0.0157, -0.0101, -0.0102], + [ 0.0052, -0.0114, -0.0042, ..., -0.0054, 0.0070, -0.0181]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.mlp.layer2.weight': tensor([[-0.0096, -0.0122, -0.0084, ..., -0.0148, -0.0040, -0.0079], + [ 0.0123, 0.0056, 0.0131, ..., 0.0037, 0.0097, -0.0077], + [ 0.0052, -0.0040, -0.0037, ..., -0.0010, -0.0015, 0.0038], + ..., + [ 0.0037, -0.0034, 0.0304, ..., -0.0075, -0.0130, 0.0051], + [ 0.0050, 0.0050, 0.0135, ..., 0.0075, 0.0049, -0.0062], + [ 0.0220, -0.0164, -0.0082, ..., 0.0068, 0.0042, 0.0048]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0432, -0.0420, -0.0415, ..., 0.0050, 0.0053, 0.0052], + [ 0.0352, 0.0339, 0.0330, ..., -0.0036, -0.0035, -0.0036], + [-0.0422, -0.0403, -0.0398, ..., 0.0050, 0.0051, 0.0050], + ..., + [-0.0476, -0.0464, -0.0449, ..., 0.0034, 0.0035, 0.0036], + [-0.0413, -0.0410, -0.0393, ..., 0.0036, 0.0039, 0.0039], + [-0.0251, -0.0240, -0.0228, ..., 0.0011, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0042, -0.0036, 0.0035, ..., 0.0033, 0.0033, 0.0034], + [ 0.0006, 0.0030, 0.0020, ..., 0.0028, 0.0019, -0.0021], + [-0.0068, 0.0038, -0.0041, ..., -0.0077, -0.0069, -0.0038], + ..., + [-0.0231, -0.0007, -0.0286, ..., -0.0308, -0.0292, 0.0007], + [-0.0117, -0.0054, -0.0129, ..., -0.0142, -0.0121, 0.0092], + [ 0.0069, -0.0063, 0.0044, ..., 0.0069, 0.0062, -0.0011]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_cross_attn.1.weight': tensor([[ 6.2500e-02, 6.0791e-02, 6.0303e-02, ..., -1.9550e-05, + -4.8876e-05, -4.3213e-06], + [ 7.8125e-03, 7.8125e-03, 7.8125e-03, ..., -3.9577e-05, + -1.1349e-04, -2.2888e-04], + [-2.0264e-02, -1.9531e-02, -1.8555e-02, ..., -6.7139e-04, + -6.7139e-04, -6.8283e-04], + ..., + [-2.6489e-02, -2.6489e-02, -2.5635e-02, ..., -3.9864e-04, + -3.5667e-04, -3.1853e-04], + [-2.7100e-02, -2.6489e-02, -2.5635e-02, ..., -2.0123e-04, + -1.5545e-04, -2.0885e-04], + [ 5.7373e-02, 5.5664e-02, 5.4932e-02, ..., -2.0409e-04, + -8.5831e-05, -2.3174e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0074, -0.0041, -0.0145, ..., 0.0043, 0.0080, -0.0095], + [ 0.0068, 0.0026, -0.0005, ..., -0.0011, -0.0033, 0.0017], + [ 0.0147, -0.0042, -0.0042, ..., 0.0043, -0.0070, 0.0095], + ..., + [ 0.0679, 0.0045, -0.0058, ..., 0.0059, -0.0175, 0.0630], + [ 0.0078, -0.0104, -0.0044, ..., -0.0155, -0.0220, 0.0183], + [-0.0065, 0.0087, -0.0153, ..., 0.0106, 0.0091, -0.0083]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_mlp.1.weight': tensor([[ 0.0096, 0.0090, 0.0082, ..., 0.0003, 0.0002, 0.0002], + [ 0.0325, 0.0317, 0.0315, ..., -0.0037, -0.0037, -0.0036], + [ 0.0457, 0.0444, 0.0439, ..., -0.0038, -0.0038, -0.0038], + ..., + [ 0.0085, 0.0082, 0.0088, ..., -0.0020, -0.0020, -0.0019], + [-0.0181, -0.0167, -0.0159, ..., 0.0015, 0.0014, 0.0014], + [ 0.0052, 0.0052, 0.0052, ..., -0.0024, -0.0025, -0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_mlp.2.weight': tensor([[-0.0024, -0.0014, -0.0033, ..., 0.0035, 0.0031, 0.0024], + [-0.0004, -0.0010, 0.0036, ..., 0.0014, 0.0003, 0.0015], + [ 0.0036, -0.0046, -0.0056, ..., -0.0038, -0.0011, -0.0038], + ..., + [-0.0011, 0.0083, 0.0131, ..., -0.0092, 0.0016, -0.0082], + [ 0.0167, 0.0129, 0.0238, ..., -0.0120, -0.0175, -0.0125], + [ 0.0157, -0.0153, -0.0156, ..., -0.0150, -0.0161, -0.0140]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.after_proj.weight': tensor([[-0.0284, 0.0079, 0.0044, ..., -0.0072, -0.0132, 0.0128], + [ 0.0086, 0.0190, 0.0042, ..., -0.0061, -0.0064, 0.0031], + [ 0.0120, -0.0156, -0.0128, ..., -0.0083, 0.0104, 0.0078], + ..., + [-0.0187, -0.0171, 0.0166, ..., -0.0065, -0.0040, -0.0048], + [-0.0098, -0.0085, 0.0060, ..., 0.0056, 0.0134, -0.0042], + [-0.0182, 0.0089, -0.0125, ..., -0.0096, -0.0066, -0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.after_proj.bias': tensor([ 0.0125, 0.0034, 0.0019, ..., 0.0067, 0.0037, -0.0056], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_proj.weight': tensor([[ 0.0243, -0.0233, -0.0165, ..., 0.0219, -0.0100, -0.0254], + [-0.0001, -0.0242, -0.0025, ..., -0.0129, 0.0074, 0.0037], + [ 0.0148, 0.0154, 0.0070, ..., 0.0044, -0.0095, 0.0009], + ..., + [-0.0036, -0.0067, -0.0034, ..., 0.0040, 0.0106, -0.0075], + [ 0.0281, 0.0144, -0.0022, ..., 0.0077, -0.0160, 0.0281], + [ 0.0074, -0.0193, 0.0089, ..., -0.0330, 0.0123, -0.0231]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_norm.weight': tensor([ 1.5625e+00, 4.9414e-01, 5.1562e-01, 3.0273e-01, 1.4922e+00, + 4.3335e-03, -1.1719e-02, -4.3392e-05, 1.8616e-03, 8.9722e-03, + 1.4827e-06, 3.5623e-08, 1.2266e+00, -6.4392e-03, -3.1876e-17, + -3.2715e-02, -1.3367e-02, -4.3213e-02, 2.4414e-03, 1.8188e-02, + 1.6895e-01, 2.8372e-05, 6.4062e-01, 1.5859e+00, 6.0547e-01, + 6.8750e-01, 4.7656e-01, 1.6699e-01, 2.0020e-02, 5.5847e-03, + 2.7710e-02, -4.9591e-05, 6.1340e-03, -2.0874e-02, 3.3691e-02, + 4.4434e-02, 1.2451e-01, 8.9722e-03, -7.6599e-03, 1.5736e-05, + -5.0545e-05, -2.2507e-04, 1.2146e-02, 1.3125e+00, 1.3184e-01, + 1.3906e+00, 6.6797e-01, 2.1094e-01, -3.7598e-02, -1.5869e-02, + 2.9297e-02, 5.1025e-02, 1.3477e-01, 3.3855e-05, 2.4080e-05, + -5.6252e-07, 4.5061e-05, -7.8678e-05, -6.1279e-02, -4.2783e-09, + 4.6875e-02, -1.9932e-04, -3.3691e-02, -3.5889e-02, 8.7500e-01, + 1.6016e+00, 1.4688e+00, 1.4531e+00, 3.1250e-01, -4.6997e-03, + 5.7068e-03, 3.9339e-05, 1.2939e-02, -4.2114e-03, -1.9932e-04, + 9.0525e-07, 3.2715e-02, 2.4780e-02, 7.6050e-15, 8.1250e-01, + 1.0078e+00, -4.6387e-03, -1.1368e-03, 3.2959e-02, 1.2988e-01, + 2.0508e-02, 1.5234e+00, 5.6641e-01, 1.3984e+00, 1.5000e+00, + 2.8125e-01, 5.1562e-01, 5.1025e-02, -1.1902e-02, 2.1606e-02, + -1.8024e-04, -2.0752e-03, -3.7231e-03, 1.8463e-03, 3.2959e-02, + 1.2207e-01, 2.5177e-04, 1.1641e+00, 1.7405e-05, 2.4915e-05, + -4.7684e-05, -8.5354e-05, 9.6875e-01, 1.5391e+00, 1.4844e+00, + 1.5469e+00, 3.7891e-01, 3.2227e-02, 4.5410e-02, 2.2827e-02, + 1.9287e-02, 1.4746e-01, 6.7711e-05, -7.6890e-06, 1.7658e-06, + 1.8120e-05, 9.2030e-05, 1.1094e+00, 2.7831e-10, 2.4536e-02, + -1.2219e-05, 7.6562e-01, 1.1172e+00], dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.self_attn.k_proj.weight': tensor([[-0.0388, 0.0049, -0.0071, ..., 0.0249, 0.0139, 0.0238], + [-0.0056, 0.0130, -0.0036, ..., -0.0114, 0.0283, -0.0051], + [-0.0294, 0.0200, -0.0027, ..., 0.0057, 0.0228, -0.0034], + ..., + [-0.0057, 0.0018, -0.0045, ..., 0.0026, 0.0011, 0.0127], + [ 0.0050, 0.0120, -0.0009, ..., -0.0046, -0.0048, 0.0089], + [ 0.0190, -0.0225, -0.0019, ..., -0.0131, 0.0058, -0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.k_norm.weight': tensor([ 1.6328e+00, 8.9844e-01, 3.9648e-01, 3.6328e-01, 1.4922e+00, + 4.3030e-03, 1.6235e-02, -4.3392e-05, 6.6528e-03, 8.9722e-03, + 3.5018e-06, -1.6065e-08, 1.2266e+00, -6.0425e-03, -8.1966e-17, + -2.6489e-02, -4.1199e-03, 1.1768e-01, 4.2725e-03, 1.9836e-03, + 2.7466e-02, -7.7724e-05, 6.5625e-01, 1.4297e+00, 2.5391e-01, + 7.0312e-01, 5.5859e-01, 2.3926e-01, -1.8677e-02, 1.3000e-02, + 2.9663e-02, 1.3947e-05, 6.5308e-03, 5.4688e-02, 3.4912e-02, + -1.8311e-02, -1.2573e-02, 8.9722e-03, 2.6489e-02, 1.5974e-05, + -8.2016e-05, -2.2030e-04, 1.2024e-02, 1.5078e+00, 9.9219e-01, + 1.3750e+00, 7.3047e-01, 2.5586e-01, 6.5308e-03, 6.7871e-02, + 3.1738e-02, 7.4158e-03, 3.6774e-03, 6.2466e-05, -1.4484e-05, + -4.8429e-07, 2.9206e-05, 4.9114e-05, 9.8877e-03, -4.5693e-09, + -2.2217e-02, -1.9741e-04, -4.4434e-02, -3.8818e-02, 7.5781e-01, + 1.4609e+00, 1.5000e+00, 1.4375e+00, 2.9688e-01, 4.4556e-03, + -1.6937e-03, 3.6716e-05, 2.1515e-03, 4.3030e-03, -1.9932e-04, + -8.9034e-07, -3.0029e-02, 2.5024e-02, 7.4940e-15, 8.1250e-01, + 1.0078e+00, 5.7617e-02, 1.6022e-03, 1.1719e-02, -3.0884e-02, + 2.0752e-02, 1.4844e+00, 9.2969e-01, 1.5156e+00, 1.5391e+00, + 4.8584e-02, 5.1172e-01, 9.9487e-03, 6.8970e-03, 1.5625e-02, + 2.0885e-04, -6.2180e-04, 4.5166e-02, 4.4250e-03, 1.0498e-02, + 1.1414e-02, -1.1492e-04, 1.1641e+00, 1.7643e-05, 2.4319e-05, + -4.7207e-05, -4.1246e-05, 6.7188e-01, 1.3672e+00, 1.3828e+00, + 1.6641e+00, 3.8086e-01, -4.3701e-02, 1.1963e-01, 2.4170e-02, + 2.2583e-02, -1.5259e-02, -1.0729e-06, -2.0266e-06, -1.4156e-06, + 4.1485e-05, -1.6212e-04, 1.1094e+00, -3.0923e-10, -1.7853e-03, + -8.9407e-06, 7.6562e-01, 1.1172e+00], dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.self_attn.v_proj.weight': tensor([[ 0.0025, -0.0052, -0.0096, ..., -0.0293, -0.0051, 0.0109], + [ 0.0068, 0.0021, -0.0122, ..., -0.0017, -0.0008, -0.0042], + [-0.0245, 0.0041, -0.0022, ..., 0.0148, 0.0055, -0.0143], + ..., + [ 0.0137, 0.0065, -0.0337, ..., -0.0454, 0.0104, -0.0117], + [ 0.0017, -0.0108, 0.0018, ..., -0.0069, -0.0074, -0.0117], + [-0.0015, 0.0028, -0.0188, ..., -0.0042, 0.0068, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.output_proj.weight': tensor([[-0.0040, 0.0315, -0.0271, ..., -0.0048, -0.0188, 0.0134], + [ 0.0038, -0.0175, -0.0282, ..., 0.0054, 0.0064, -0.0032], + [-0.0034, -0.0038, 0.0041, ..., 0.0038, -0.0100, 0.0074], + ..., + [ 0.0094, -0.0041, -0.0128, ..., 0.0277, 0.0123, -0.0049], + [ 0.0029, 0.0190, -0.0044, ..., 0.0171, 0.0071, -0.0211], + [-0.0037, -0.0123, 0.0038, ..., -0.0239, 0.0061, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_proj.weight': tensor([[ 0.0055, -0.0037, 0.0042, ..., -0.0042, 0.0129, 0.0038], + [ 0.0061, 0.0057, 0.0023, ..., 0.0026, 0.0038, -0.0031], + [ 0.0036, -0.0056, 0.0003, ..., -0.0069, 0.0186, -0.0042], + ..., + [-0.0025, 0.0046, 0.0038, ..., 0.0108, 0.0037, 0.0046], + [-0.0059, -0.0049, 0.0035, ..., 0.0124, 0.0095, 0.0079], + [ 0.0080, 0.0013, -0.0031, ..., -0.0109, 0.0103, -0.0040]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_norm.weight': tensor([ 6.7749e-03, 1.2305e-01, 2.2339e-02, 6.0156e-01, 2.3315e-02, + 9.6436e-03, 2.0386e-02, 4.0588e-03, 1.5137e-02, 1.0645e-01, + 2.2583e-02, 4.7852e-02, 2.6172e-01, 2.6489e-02, 1.7969e-01, + 1.8433e-02, 4.6875e-02, 2.4048e-02, 5.5847e-03, 8.3008e-03, + 1.2268e-02, 1.5918e-01, 2.7148e-01, 1.9238e-01, 2.1118e-02, + 1.4941e-01, 4.4922e-02, 8.5449e-03, 4.0820e-01, 1.3611e-02, + 6.3782e-03, 5.0391e-01, 4.4922e-02, 4.1748e-02, 2.0752e-02, + 8.9722e-03, 7.5684e-02, 5.2344e-01, 5.9375e-01, 3.5400e-03, + 1.3062e-02, 1.0156e-01, 8.3618e-03, 5.5859e-01, 1.6479e-02, + 3.2715e-02, 8.4839e-03, 4.2773e-01, 4.2419e-03, 2.1484e-02, + 6.3477e-02, 2.2949e-02, 1.7334e-02, 5.6641e-01, -7.0095e-05, + 3.2227e-02, 3.3691e-02, 6.7383e-02, 1.5198e-02, 1.5198e-02, + 2.5330e-03, 1.2329e-02, 1.5717e-03, 3.5547e-01, 1.4648e-02, + -1.2684e-04, 1.8433e-02, 5.8350e-02, 7.4707e-02, 9.2163e-03, + 9.2773e-03, 6.3672e-01, 1.9789e-05, 6.0938e-01, 1.5381e-02, + 6.7139e-04, 9.5825e-03, 2.9297e-01, 2.9907e-02, 2.9564e-04, + 9.3262e-02, 8.3008e-03, 4.1748e-02, 7.6294e-03, 2.4170e-02, + 5.2490e-03, 2.2827e-02, 6.4941e-02, 2.0898e-01, 9.8877e-03, + 5.9814e-03, 1.3489e-02, 4.0527e-02, 2.0752e-02, 3.5156e-02, + 2.5269e-02, 9.3384e-03, 2.4414e-02, 5.0293e-02, 1.9531e-02, + 2.9907e-02, 1.5625e-02, -3.8300e-03, 5.1514e-02, 6.1719e-01, + 5.2734e-02, 1.1749e-03, 4.9219e-01, 1.8799e-02, 2.3315e-02, + 1.1902e-02, 2.0294e-03, -6.1512e-05, 8.0078e-02, 4.6094e-01, + 1.4099e-02, 2.3682e-02, 6.1328e-01, 1.1084e-01, 3.0518e-02, + 1.1169e-02, 6.9922e-01, 1.4453e-01, 1.3916e-02, 7.8125e-03, + 2.4414e-03, 1.8921e-02, 1.1520e-03], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_proj.weight': tensor([[ 0.0038, -0.0034, 0.0050, ..., -0.0018, -0.0112, 0.0114], + [ 0.0039, 0.0003, -0.0025, ..., 0.0014, -0.0126, -0.0041], + [ 0.0024, -0.0016, -0.0035, ..., 0.0120, -0.0132, -0.0042], + ..., + [ 0.0052, 0.0046, -0.0034, ..., 0.0023, 0.0061, -0.0125], + [ 0.0093, 0.0029, -0.0020, ..., -0.0042, 0.0042, 0.0027], + [-0.0091, 0.0007, -0.0101, ..., -0.0042, -0.0160, 0.0046]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_norm.weight': tensor([ 6.7749e-03, 1.2305e-01, 2.2461e-02, 6.0156e-01, 2.3193e-02, + 9.4604e-03, 2.0386e-02, 4.0588e-03, 1.5381e-02, 1.0645e-01, + 2.2583e-02, 4.6143e-02, 2.6172e-01, 2.6611e-02, 1.7969e-01, + 1.9409e-02, 4.6387e-02, 2.4048e-02, 5.5237e-03, 8.1787e-03, + 1.2390e-02, 1.6016e-01, 2.7344e-01, 1.9238e-01, 2.0752e-02, + 1.4941e-01, 4.3945e-02, 6.3477e-03, 4.1016e-01, 1.3489e-02, + 6.6528e-03, 5.0391e-01, 4.4922e-02, 4.1992e-02, 2.1118e-02, + 8.3008e-03, 7.4707e-02, 5.2344e-01, 5.9375e-01, 3.6011e-03, + 1.3062e-02, 1.0107e-01, 8.3618e-03, 5.5859e-01, 1.6846e-02, + 3.2959e-02, 8.7891e-03, 4.2773e-01, 4.3335e-03, 2.1362e-02, + 6.3477e-02, 2.3193e-02, 1.7334e-02, 5.6641e-01, -7.0095e-05, + 3.2227e-02, 3.3203e-02, 6.7383e-02, 1.5259e-02, 1.3672e-02, + 2.4261e-03, 1.2268e-02, 2.3956e-03, 3.5547e-01, 1.4221e-02, + 2.9755e-04, 1.7822e-02, 5.8350e-02, 7.4707e-02, 9.3384e-03, + 9.5825e-03, 6.3672e-01, 1.9789e-05, 6.0938e-01, 1.5625e-02, + -5.8365e-04, 9.6436e-03, 2.9297e-01, 3.0640e-02, 3.8910e-04, + 9.3750e-02, 8.1177e-03, 4.0771e-02, 7.3547e-03, 2.3682e-02, + 5.2490e-03, 2.3193e-02, 6.5430e-02, 2.0898e-01, 9.8877e-03, + 6.0120e-03, 1.2329e-02, 4.0527e-02, 2.0996e-02, 3.6133e-02, + 2.5269e-02, 9.3994e-03, 2.4414e-02, 5.0293e-02, 1.9409e-02, + 2.9175e-02, 1.5747e-02, 3.8300e-03, 5.2246e-02, 6.1719e-01, + 5.2734e-02, 1.1444e-03, 4.9219e-01, 1.8433e-02, 2.3071e-02, + 1.2695e-02, 2.0294e-03, -5.4836e-05, 8.0078e-02, 4.6094e-01, + 1.3672e-02, 2.3438e-02, 6.1328e-01, 1.1182e-01, 3.0640e-02, + 1.0864e-02, 7.0312e-01, 1.4355e-01, 1.2756e-02, 7.5378e-03, + 2.3346e-03, 1.8677e-02, -9.3460e-04], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.v_proj.weight': tensor([[ 0.0133, -0.0066, -0.0176, ..., -0.0031, -0.0162, -0.0116], + [-0.0157, 0.0014, -0.0052, ..., -0.0026, -0.0210, 0.0210], + [-0.0116, -0.0044, -0.0012, ..., -0.0032, 0.0299, 0.0249], + ..., + [-0.0052, 0.0014, -0.0125, ..., 0.0035, 0.0112, -0.0006], + [ 0.0284, 0.0068, 0.0254, ..., 0.0093, 0.0096, 0.0153], + [ 0.0162, 0.0038, 0.0074, ..., 0.0071, -0.0186, -0.0244]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.output_proj.weight': tensor([[ 0.0153, 0.0042, -0.0161, ..., 0.0133, 0.0081, 0.0150], + [ 0.0088, -0.0010, -0.0029, ..., 0.0084, 0.0030, 0.0031], + [-0.0039, -0.0069, 0.0044, ..., -0.0048, 0.0067, -0.0067], + ..., + [ 0.0047, -0.0082, 0.0043, ..., 0.0047, -0.0096, 0.0056], + [ 0.0045, -0.0133, 0.0067, ..., -0.0403, -0.0249, -0.0233], + [-0.0021, -0.0025, 0.0154, ..., 0.0086, 0.0036, 0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0068, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [ 0.0038, 0.0150, -0.0159, ..., -0.0034, -0.0056, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0068, -0.0057], + [-0.0337, -0.0031, 0.0048, ..., -0.0032, 0.0172, 0.0231], + [-0.0255, 0.0004, -0.0044, ..., 0.0408, 0.0430, -0.0120]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0033], + [ 0.0198, -0.0208, 0.0129, ..., 0.0089, -0.0337, -0.0051], + [ 0.0302, 0.0101, -0.0018, ..., -0.0304, 0.0199, -0.0039], + ..., + [ 0.0171, -0.0276, -0.0066, ..., 0.0231, 0.0182, 0.0011], + [ 0.0221, 0.0388, 0.0047, ..., 0.0042, -0.0488, -0.0361], + [ 0.0045, 0.0118, -0.0059, ..., 0.0010, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img.weight': tensor([[ 0.0273, -0.0041, 0.0118, ..., 0.0022, 0.0347, 0.0615], + [-0.0258, -0.0047, -0.0197, ..., 0.0139, -0.0042, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0097, 0.0132, 0.0184, ..., 0.0110, 0.0089, 0.0515], + [-0.0017, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.mlp.layer1.weight': tensor([[-0.0108, 0.0063, 0.0072, ..., -0.0044, -0.0053, -0.0164], + [-0.0096, 0.0017, -0.0006, ..., -0.0088, 0.0035, -0.0081], + [-0.0038, 0.0062, 0.0038, ..., -0.0022, -0.0123, -0.0510], + ..., + [-0.0118, 0.0111, 0.0038, ..., 0.0068, -0.0043, -0.0075], + [-0.0062, 0.0075, 0.0042, ..., -0.0066, -0.0088, -0.0383], + [-0.0030, 0.0275, -0.0038, ..., 0.0016, 0.0046, -0.0094]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.mlp.layer2.weight': tensor([[-0.0107, -0.0052, -0.0210, ..., 0.0140, -0.0079, -0.0112], + [-0.0051, 0.0047, -0.0121, ..., -0.0032, 0.0078, -0.0229], + [-0.0051, 0.0137, -0.0172, ..., 0.0005, 0.0127, -0.0106], + ..., + [-0.0009, 0.0081, -0.0381, ..., 0.0082, 0.0142, -0.0155], + [-0.0127, 0.0184, 0.0122, ..., 0.0067, 0.0096, -0.0037], + [-0.0101, -0.0040, -0.0364, ..., -0.0003, 0.0109, -0.0334]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0289, 0.0291, 0.0275, ..., -0.0041, -0.0042, -0.0043], + [-0.0254, -0.0244, -0.0242, ..., 0.0043, 0.0047, 0.0046], + [-0.0297, -0.0293, -0.0286, ..., 0.0045, 0.0048, 0.0048], + ..., + [-0.0322, -0.0312, -0.0305, ..., 0.0021, 0.0022, 0.0023], + [-0.0327, -0.0325, -0.0317, ..., 0.0020, 0.0021, 0.0021], + [-0.0046, -0.0046, -0.0045, ..., 0.0001, 0.0001, 0.0001]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0049, -0.0042, -0.0014, ..., -0.0048, -0.0051, -0.0014], + [ 0.0017, -0.0020, -0.0017, ..., -0.0018, -0.0022, 0.0031], + [ 0.0099, -0.0096, -0.0102, ..., -0.0041, -0.0034, 0.0007], + ..., + [ 0.0388, -0.0427, -0.0432, ..., -0.0249, -0.0217, 0.0142], + [-0.0041, 0.0042, -0.0049, ..., 0.0044, 0.0045, 0.0060], + [-0.0148, 0.0173, 0.0148, ..., 0.0162, 0.0146, 0.0165]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_cross_attn.1.weight': tensor([[ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., -1.3065e-04, + -6.0320e-05, -2.1267e-04], + [ 6.1768e-02, 6.0303e-02, 5.8594e-02, ..., 3.3569e-04, + 4.4632e-04, 4.7874e-04], + [-5.3955e-02, -5.2490e-02, -5.1025e-02, ..., -5.2154e-06, + 5.0545e-05, -5.7459e-05], + ..., + [ 3.3447e-02, 3.2715e-02, 3.1982e-02, ..., 1.0204e-04, + 3.3617e-05, 6.2943e-05], + [ 5.9570e-02, 5.8105e-02, 5.7617e-02, ..., 5.2261e-04, + 5.6458e-04, 5.5695e-04], + [-4.8828e-02, -4.8096e-02, -4.7363e-02, ..., -3.2997e-04, + -3.7384e-04, -3.3379e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0254, 0.0063, -0.0156, ..., 0.0030, -0.0043, -0.0016], + [-0.0216, 0.0137, 0.0175, ..., -0.0090, -0.0137, 0.0099], + [ 0.0098, 0.0008, -0.0087, ..., 0.0060, 0.0008, -0.0008], + ..., + [ 0.0225, 0.0156, -0.0128, ..., 0.0072, 0.0080, 0.0045], + [ 0.0117, 0.0444, -0.0057, ..., 0.0040, 0.0225, -0.0079], + [-0.0187, 0.0161, 0.0192, ..., -0.0135, -0.0082, 0.0210]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_mlp.1.weight': tensor([[-0.0571, -0.0557, -0.0542, ..., 0.0029, 0.0031, 0.0031], + [-0.0259, -0.0244, -0.0232, ..., 0.0008, 0.0008, 0.0007], + [ 0.0344, 0.0332, 0.0332, ..., -0.0016, -0.0016, -0.0016], + ..., + [ 0.0304, 0.0294, 0.0295, ..., -0.0018, -0.0017, -0.0015], + [ 0.0356, 0.0342, 0.0327, ..., -0.0010, -0.0009, -0.0010], + [-0.0295, -0.0288, -0.0282, ..., 0.0021, 0.0021, 0.0020]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_mlp.2.weight': tensor([[-0.0233, -0.0145, 0.0118, ..., 0.0041, 0.0162, -0.0124], + [-0.0010, -0.0058, -0.0026, ..., -0.0016, 0.0057, 0.0014], + [-0.0004, -0.0009, 0.0005, ..., -0.0008, 0.0005, 0.0001], + ..., + [ 0.0432, -0.0041, -0.0190, ..., -0.0137, 0.0040, 0.0197], + [ 0.0226, -0.0126, -0.0173, ..., -0.0151, 0.0124, 0.0167], + [ 0.0072, -0.0306, -0.0190, ..., -0.0220, 0.0315, 0.0182]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.after_proj.weight': tensor([[-0.0034, 0.0076, -0.0066, ..., -0.0079, 0.0073, 0.0118], + [ 0.0095, -0.0077, 0.0047, ..., -0.0018, 0.0079, 0.0058], + [ 0.0027, 0.0077, 0.0167, ..., 0.0063, 0.0069, 0.0176], + ..., + [ 0.0086, -0.0040, 0.0145, ..., 0.0044, -0.0072, 0.0212], + [ 0.0033, -0.0074, 0.0148, ..., -0.0079, -0.0046, -0.0078], + [-0.0034, 0.0047, -0.0098, ..., -0.0081, 0.0216, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.after_proj.bias': tensor([-0.0038, -0.0012, 0.0049, ..., -0.0038, 0.0101, -0.0013], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_proj.weight': tensor([[-0.0121, 0.0339, 0.0119, ..., -0.0087, 0.0035, -0.0254], + [-0.0052, 0.0046, 0.0090, ..., 0.0026, 0.0165, -0.0212], + [ 0.0130, 0.0161, 0.0101, ..., -0.0052, -0.0282, 0.0151], + ..., + [-0.0036, -0.0046, 0.0041, ..., 0.0019, 0.0104, 0.0014], + [-0.0120, -0.0302, 0.0251, ..., -0.0388, 0.0052, -0.0098], + [-0.0039, -0.0037, 0.0004, ..., 0.0075, 0.0177, 0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_norm.weight': tensor([ 9.3750e-01, 1.7422e+00, 1.4844e+00, 1.5547e+00, 1.5430e-01, + 1.2207e-01, 1.7212e-02, -1.6065e-08, -2.3954e-33, 3.5553e-03, + 1.1353e-02, 1.0791e-01, -5.8594e-03, -4.5166e-03, 1.6022e-03, + -1.0490e-04, 6.5430e-02, 1.0391e+00, 1.1094e+00, 7.7820e-04, + 4.1485e-05, 1.2891e+00, 2.0312e+00, 1.8516e+00, 1.0625e+00, + 8.8281e-01, 1.3750e+00, -1.2817e-02, -6.9046e-04, 3.4714e-04, + -2.3041e-03, -2.5940e-04, 3.7354e-02, 1.1641e+00, 6.1035e-03, + 1.1172e+00, 1.6113e-02, 3.1710e-05, 1.4221e-02, -7.5378e-03, + 2.0862e-05, 1.6212e-05, 1.5020e-05, 2.0938e+00, 1.3516e+00, + 1.4922e+00, 1.5234e+00, 1.4688e+00, 9.4238e-02, 5.4932e-04, + 3.1982e-02, 2.7148e-01, 1.1719e+00, 1.1475e-01, 2.1362e-03, + 8.2520e-02, 2.8839e-03, 1.0107e-01, -4.4107e-05, 3.6163e-03, + -1.1978e-03, 5.2185e-03, -5.2490e-03, 1.1719e+00, 1.9375e+00, + 6.9531e-01, 8.5547e-01, 1.3828e+00, 4.5410e-02, -2.3193e-02, + 1.1475e-02, 3.3993e-08, -3.8037e-33, 9.4604e-03, 1.8921e-03, + -3.1586e-03, -1.5450e-04, 2.0386e-02, -4.1748e-02, 2.5635e-03, + 4.6387e-02, 4.7607e-03, 1.8188e-02, 8.7738e-04, 9.6321e-05, + -5.2185e-03, 8.4375e-01, 8.2031e-01, 1.5312e+00, 1.4609e+00, + 7.6562e-01, -5.9814e-03, -2.0885e-04, -5.9605e-05, -2.3651e-03, + 2.6894e-04, 2.1973e-02, 1.3770e-01, 2.3804e-03, 4.5654e-02, + -2.1362e-02, -8.4639e-06, 1.2891e-01, 6.4850e-05, 5.7983e-03, + -1.5736e-05, -2.9206e-05, 5.1953e-01, 1.7266e+00, 1.4922e+00, + 8.3984e-01, 7.5781e-01, 9.4727e-02, 4.0054e-04, 2.5391e-02, + 1.1562e+00, 8.0078e-02, 1.1875e+00, -7.3910e-05, 9.0332e-03, + 1.9043e-02, 4.7852e-02, 2.4438e-05, -2.1210e-03, 2.1648e-04, + -2.1729e-02, 1.1328e+00, -9.9487e-03], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.self_attn.k_proj.weight': tensor([[-1.5747e-02, 6.3171e-03, 2.8381e-03, ..., -1.1230e-02, + -3.0212e-03, 3.7842e-03], + [-1.0864e-02, -2.5635e-02, 1.5503e-02, ..., -7.3242e-03, + -4.9438e-03, -5.3406e-03], + [ 5.0964e-03, 4.3945e-03, 9.0942e-03, ..., 1.7090e-02, + -6.8665e-03, 3.7384e-03], + ..., + [-4.6082e-03, 4.9133e-03, 4.7913e-03, ..., 1.9836e-03, + -3.8605e-03, 3.3417e-03], + [-1.3550e-02, -1.5736e-05, 2.5757e-02, ..., -1.5381e-02, + 7.1716e-03, 5.3406e-03], + [ 1.9226e-03, 4.5166e-03, -7.6904e-03, ..., -4.6692e-03, + 6.8665e-03, 2.1362e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.k_norm.weight': tensor([ 8.7891e-01, 1.7109e+00, 1.5000e+00, 1.4922e+00, 1.5625e-01, + 1.2793e-01, 1.1902e-02, -6.7521e-09, -1.7935e-33, -1.6251e-03, + 1.1169e-02, 1.0889e-01, -1.0925e-02, 3.1006e-02, -2.0142e-03, + 9.8705e-05, 1.3123e-02, 1.0391e+00, 1.1094e+00, 7.6675e-04, + 7.2479e-05, 1.2891e+00, 1.7969e+00, 1.7188e+00, 1.2891e-01, + 3.8281e-01, 1.4219e+00, -1.5015e-02, 3.2425e-04, -3.0136e-04, + -2.4567e-03, 2.9564e-04, -3.6133e-02, 1.1641e+00, 2.5787e-03, + 1.1094e+00, 1.6113e-02, -3.0518e-05, -1.4954e-02, -7.6599e-03, + -1.0788e-05, 2.3365e-05, -5.0306e-05, 1.6016e+00, 6.2109e-01, + 1.4531e+00, 1.5781e+00, 1.4922e+00, -5.6885e-02, 5.0735e-04, + 1.9653e-02, -9.3994e-03, 1.1719e+00, 1.2354e-01, 2.1515e-03, + 8.2520e-02, -3.9368e-03, 1.0156e-01, -3.3855e-05, 7.0801e-02, + -1.2054e-03, 2.3315e-02, -1.5869e-02, 1.1719e+00, 1.9609e+00, + 9.9219e-01, 7.5781e-01, 1.3672e+00, 2.6855e-02, -1.0193e-02, + -2.4170e-02, 9.8255e-08, 4.4778e-33, 1.0010e-02, 1.4587e-02, + 4.3945e-02, 1.6403e-04, 3.7842e-02, -5.2490e-02, -2.6093e-03, + -4.6692e-03, 1.9287e-02, 1.5198e-02, 8.6975e-04, 9.9182e-05, + 3.2043e-03, 1.3828e+00, 1.3281e+00, 1.6250e+00, 1.5156e+00, + 3.1445e-01, 2.5177e-03, -6.5231e-04, -1.4687e-04, 2.4109e-03, + 3.3951e-04, 4.9744e-03, -2.4170e-02, -1.9684e-03, 7.1411e-03, + 2.1484e-02, -3.4332e-05, 1.2793e-01, -2.4080e-05, 5.7678e-03, + 1.9908e-05, -2.6584e-05, 1.4219e+00, 1.9531e+00, 1.4609e+00, + 1.3086e-01, 7.0703e-01, 5.0659e-03, 4.2915e-04, 4.6082e-03, + 1.1797e+00, -4.9438e-03, 1.1875e+00, 5.9366e-05, -1.8555e-02, + 1.9409e-02, 4.6387e-02, 3.0398e-05, 4.4189e-02, 2.1935e-04, + 2.8198e-02, 1.1328e+00, -1.0071e-02], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.self_attn.v_proj.weight': tensor([[-0.0090, -0.0089, 0.0096, ..., 0.0508, -0.0099, -0.0027], + [-0.0032, 0.0112, -0.0091, ..., 0.0052, -0.0070, -0.0080], + [-0.0040, 0.0182, -0.0139, ..., -0.0025, 0.0050, 0.0128], + ..., + [ 0.0008, 0.0008, -0.0097, ..., -0.0038, 0.0085, 0.0115], + [ 0.0040, 0.0105, -0.0109, ..., 0.0032, -0.0037, -0.0028], + [-0.0021, -0.0089, 0.0081, ..., 0.0151, 0.0078, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.output_proj.weight': tensor([[ 0.0198, 0.0126, 0.0085, ..., 0.0065, 0.0132, -0.0049], + [ 0.0198, -0.0013, 0.0054, ..., -0.0305, -0.0049, 0.0111], + [-0.0098, 0.0184, -0.0212, ..., -0.0133, -0.0063, 0.0136], + ..., + [-0.0292, -0.0199, 0.0059, ..., -0.0051, -0.0112, 0.0065], + [-0.0072, 0.0178, -0.0261, ..., -0.0131, 0.0439, 0.0035], + [-0.0305, -0.0042, 0.0276, ..., 0.0226, -0.0085, -0.0272]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_proj.weight': tensor([[-0.0057, -0.0057, 0.0170, ..., -0.0219, -0.0225, 0.0112], + [-0.0082, 0.0065, 0.0289, ..., -0.0229, -0.0029, 0.0105], + [ 0.0126, -0.0037, -0.0076, ..., 0.0052, 0.0110, -0.0043], + ..., + [ 0.0104, 0.0056, -0.0078, ..., -0.0189, -0.0057, 0.0044], + [ 0.0064, 0.0035, -0.0112, ..., 0.0045, -0.0034, 0.0016], + [-0.0064, -0.0037, 0.0124, ..., 0.0123, -0.0010, 0.0044]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_norm.weight': tensor([0.0967, 0.0156, 0.4414, 0.0576, 0.1416, 0.0320, 0.1377, 0.1201, 0.5195, + 0.0297, 0.0488, 0.0168, 0.0378, 0.0713, 0.0508, 0.1836, 0.4746, 0.0219, + 0.7578, 0.0405, 0.0283, 0.1533, 0.0262, 0.0757, 0.0425, 0.3906, 0.5156, + 0.2617, 0.0684, 0.0405, 0.0231, 0.2178, 0.6797, 0.2402, 0.0322, 0.0349, + 0.0141, 0.0928, 0.0175, 0.0359, 0.0491, 0.0986, 0.0206, 0.0298, 0.2041, + 0.0337, 0.3828, 0.0153, 0.0413, 0.0170, 0.0173, 0.4180, 0.4668, 0.3418, + 0.0708, 0.6211, 0.5977, 0.0447, 0.1060, 0.0198, 0.0767, 0.4648, 0.0071, + 0.5000, 0.1475, 0.3926, 0.0815, 0.0471, 0.0630, 0.4473, 0.0280, 0.4707, + 0.0135, 0.0366, 0.1377, 0.0698, 0.0148, 0.0991, 0.6523, 0.0337, 0.0732, + 0.0908, 0.0388, 0.0503, 0.4492, 0.0601, 0.0640, 0.0879, 0.0189, 0.1396, + 0.0327, 0.1758, 0.0408, 0.0549, 0.0376, 0.0996, 0.0088, 0.3340, 0.5195, + 0.0344, 0.0520, 0.2949, 0.0322, 0.1108, 0.0325, 0.2158, 0.0123, 0.6406, + 0.6211, 0.0175, 0.0161, 0.7188, 0.0410, 0.0160, 0.4805, 0.2090, 0.2695, + 0.6992, 0.0332, 0.0762, 0.0664, 0.5586, 0.6133, 0.0234, 0.0225, 0.1855, + 0.0366, 0.0369], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_proj.weight': tensor([[-3.1586e-03, 1.8158e-03, 4.1504e-03, ..., -1.4114e-03, + 1.0681e-02, 3.1281e-03], + [ 2.8076e-03, -6.7711e-05, -1.6861e-03, ..., 4.7913e-03, + -1.1108e-02, 5.2795e-03], + [-1.9043e-02, 7.6904e-03, -3.2043e-04, ..., 1.8433e-02, + 1.0681e-02, 3.5553e-03], + ..., + [ 5.8594e-03, 3.1128e-03, -5.7983e-03, ..., 9.4986e-04, + -1.9287e-02, 5.4932e-04], + [-9.5825e-03, -1.8539e-03, -1.0681e-02, ..., 6.0425e-03, + -1.0071e-02, 5.9509e-03], + [ 2.0264e-02, -1.4038e-03, 1.0681e-02, ..., -8.6670e-03, + 7.0190e-03, -6.2866e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_norm.weight': tensor([0.0967, 0.0157, 0.4414, 0.0574, 0.1416, 0.0327, 0.1387, 0.1206, 0.5195, + 0.0302, 0.0498, 0.0168, 0.0378, 0.0713, 0.0508, 0.1855, 0.4746, 0.0217, + 0.7539, 0.0403, 0.0281, 0.1533, 0.0264, 0.0752, 0.0422, 0.3867, 0.5156, + 0.2598, 0.0688, 0.0408, 0.0238, 0.2197, 0.6797, 0.2402, 0.0330, 0.0349, + 0.0137, 0.0928, 0.0175, 0.0354, 0.0483, 0.0996, 0.0210, 0.0303, 0.2051, + 0.0337, 0.3828, 0.0154, 0.0410, 0.0166, 0.0179, 0.4180, 0.4668, 0.3418, + 0.0708, 0.6211, 0.6016, 0.0439, 0.1050, 0.0201, 0.0767, 0.4648, 0.0075, + 0.5000, 0.1465, 0.3926, 0.0815, 0.0471, 0.0630, 0.4473, 0.0282, 0.4707, + 0.0132, 0.0366, 0.1377, 0.0693, 0.0148, 0.0996, 0.6523, 0.0334, 0.0723, + 0.0908, 0.0391, 0.0503, 0.4492, 0.0603, 0.0630, 0.0874, 0.0194, 0.1416, + 0.0337, 0.1748, 0.0408, 0.0552, 0.0378, 0.1006, 0.0087, 0.3340, 0.5195, + 0.0337, 0.0525, 0.2949, 0.0325, 0.1108, 0.0325, 0.2158, 0.0122, 0.6406, + 0.6211, 0.0172, 0.0162, 0.7188, 0.0408, 0.0161, 0.4824, 0.2090, 0.2695, + 0.6953, 0.0327, 0.0747, 0.0664, 0.5586, 0.6133, 0.0233, 0.0225, 0.1855, + 0.0378, 0.0374], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.v_proj.weight': tensor([[ 0.0226, 0.0038, -0.0243, ..., 0.0069, -0.0026, -0.0189], + [-0.0203, 0.0140, 0.0005, ..., 0.0066, -0.0025, -0.0311], + [ 0.0251, -0.0085, -0.0001, ..., -0.0010, -0.0099, -0.0452], + ..., + [ 0.0023, -0.0017, 0.0232, ..., -0.0073, -0.0212, 0.0034], + [-0.0327, -0.0032, -0.0256, ..., -0.0060, 0.0122, -0.0121], + [-0.0164, 0.0014, 0.0045, ..., -0.0046, -0.0042, 0.0089]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.output_proj.weight': tensor([[ 0.0017, 0.0038, -0.0017, ..., -0.0120, 0.0074, 0.0096], + [-0.0315, -0.0042, -0.0132, ..., -0.0078, -0.0199, -0.0077], + [ 0.0161, -0.0069, 0.0054, ..., -0.0117, 0.0118, -0.0172], + ..., + [ 0.0181, -0.0146, 0.0189, ..., 0.0119, -0.0060, -0.0081], + [-0.0137, -0.0216, 0.0060, ..., 0.0129, -0.0192, 0.0261], + [ 0.0005, -0.0215, 0.0315, ..., -0.0364, 0.0332, 0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0057, -0.0065, ..., 0.0266, -0.0278, 0.0075], + ..., + [-0.0217, -0.0410, 0.0488, ..., -0.0039, -0.0003, -0.0255], + [ 0.0036, 0.0066, 0.0013, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0059, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0101, 0.0038, ..., -0.0236, -0.0109, -0.0248], + [-0.0079, -0.0013, 0.0293, ..., 0.0072, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0143, 0.0126, -0.0094], + ..., + [-0.0035, 0.0320, -0.0233, ..., -0.0060, -0.0015, 0.0188], + [ 0.0136, -0.0352, 0.0267, ..., -0.0305, -0.0045, 0.0037], + [-0.0205, 0.0140, 0.0025, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img.weight': tensor([[-0.0079, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0029, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0057, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0018], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, -0.0027, -0.0020], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0023, ..., -0.0103, 0.0074, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.mlp.layer1.weight': tensor([[ 0.0151, -0.0038, 0.0084, ..., 0.0068, 0.0133, -0.0104], + [-0.0098, 0.0107, 0.0242, ..., 0.0029, 0.0115, -0.0012], + [-0.0037, 0.0354, -0.0305, ..., 0.0278, 0.0125, 0.0047], + ..., + [-0.0060, 0.0233, 0.0693, ..., -0.0088, 0.0012, -0.0090], + [-0.0055, -0.0082, -0.0050, ..., -0.0095, -0.0016, -0.0164], + [-0.0005, 0.0047, -0.0097, ..., -0.0092, -0.0284, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.mlp.layer2.weight': tensor([[ 0.0067, 0.0117, 0.0032, ..., -0.0220, -0.0221, -0.0071], + [ 0.0104, 0.0247, 0.0068, ..., -0.0092, 0.0356, -0.0181], + [ 0.0040, -0.0050, -0.0037, ..., 0.0150, 0.0275, 0.0042], + ..., + [ 0.0045, 0.0088, 0.0064, ..., -0.0083, -0.0062, 0.0048], + [ 0.0051, -0.0092, -0.0031, ..., 0.0077, 0.0183, 0.0037], + [-0.0036, 0.0048, 0.0170, ..., 0.0052, 0.0349, 0.0073]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0226, -0.0222, -0.0220, ..., 0.0010, 0.0008, 0.0009], + [-0.0315, -0.0306, -0.0298, ..., 0.0026, 0.0027, 0.0027], + [-0.0222, -0.0214, -0.0205, ..., 0.0021, 0.0023, 0.0023], + ..., + [ 0.0605, 0.0588, 0.0569, ..., -0.0037, -0.0037, -0.0037], + [ 0.0496, 0.0479, 0.0464, ..., -0.0032, -0.0031, -0.0030], + [-0.0256, -0.0250, -0.0248, ..., 0.0022, 0.0023, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0061, 0.0044, 0.0062, ..., 0.0099, 0.0130, 0.0059], + [ 0.0005, 0.0014, 0.0015, ..., -0.0015, 0.0004, 0.0016], + [ 0.0015, -0.0013, 0.0005, ..., 0.0037, 0.0038, -0.0014], + ..., + [-0.0237, -0.0762, -0.0615, ..., 0.0369, 0.0354, -0.0693], + [ 0.0089, 0.0082, 0.0085, ..., 0.0182, 0.0150, 0.0092], + [ 0.0164, 0.0099, 0.0124, ..., 0.0187, 0.0208, 0.0113]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_cross_attn.1.weight': tensor([[-8.2031e-02, -8.0078e-02, -7.7637e-02, ..., -1.0443e-04, + -1.2016e-04, -1.0967e-04], + [ 3.0029e-02, 2.9419e-02, 2.8564e-02, ..., 5.6624e-06, + 3.1948e-05, 1.5163e-04], + [ 3.2715e-02, 3.1494e-02, 3.0518e-02, ..., 2.6703e-04, + 2.8992e-04, 1.9836e-04], + ..., + [ 5.5908e-02, 5.4199e-02, 5.3223e-02, ..., -1.8978e-04, + -1.8501e-04, -1.1683e-04], + [ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.8311e-04, + -1.2112e-04, -1.7071e-04], + [-3.0884e-02, -2.9907e-02, -2.9175e-02, ..., -5.8889e-05, + 9.3460e-05, 1.1492e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_cross_attn.2.weight': tensor([[-7.9346e-03, 3.7842e-03, -8.8501e-03, ..., 2.8320e-02, + 3.0396e-02, 3.6926e-03], + [ 6.7139e-03, -9.3994e-03, 1.1841e-02, ..., -2.4605e-04, + 4.5166e-03, 4.6387e-03], + [ 2.0905e-03, -1.2207e-03, 4.6997e-03, ..., -7.9346e-03, + -6.3477e-03, -4.7913e-03], + ..., + [-1.5991e-02, 4.6387e-03, 3.5400e-02, ..., 2.5146e-02, + 3.2227e-02, -8.7891e-03], + [-2.3315e-02, 1.2085e-02, 3.6865e-02, ..., 1.0254e-02, + 1.5869e-02, -1.4099e-02], + [ 1.3977e-02, -8.5449e-03, -7.0801e-03, ..., 1.5717e-03, + 1.6098e-03, -4.5300e-05]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_mlp.1.weight': tensor([[-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., 1.3638e-04, + 2.5940e-04, 2.0313e-04], + [ 1.3245e-02, 1.1841e-02, 1.1597e-02, ..., 3.7193e-04, + 4.5204e-04, 4.4441e-04], + [-7.3242e-02, -7.1777e-02, -7.0801e-02, ..., -4.4823e-04, + -3.7766e-04, -4.1199e-04], + ..., + [-6.5430e-02, -6.4453e-02, -6.3477e-02, ..., 1.6785e-04, + 2.7084e-04, 1.6785e-04], + [-4.4922e-02, -4.4434e-02, -4.4189e-02, ..., 2.0981e-04, + 6.8188e-05, 7.9632e-05], + [ 6.2012e-02, 6.1035e-02, 6.0791e-02, ..., -1.4210e-04, + -1.6975e-04, -1.3351e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_mlp.2.weight': tensor([[-0.0141, 0.0030, -0.0100, ..., -0.0164, -0.0199, 0.0178], + [-0.0017, -0.0063, 0.0064, ..., -0.0015, 0.0049, -0.0028], + [-0.0016, -0.0029, -0.0036, ..., -0.0010, -0.0017, 0.0016], + ..., + [-0.0493, -0.0312, -0.0461, ..., -0.0361, -0.0109, 0.0266], + [ 0.0075, -0.0177, 0.0260, ..., 0.0075, 0.0052, -0.0075], + [ 0.0135, 0.0118, 0.0070, ..., 0.0121, 0.0072, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.after_proj.weight': tensor([[ 2.8229e-03, 1.6724e-02, -1.8311e-02, ..., 7.8125e-03, + 3.9368e-03, -6.8970e-03], + [-8.4229e-03, -3.6316e-03, 7.6599e-03, ..., -9.5844e-05, + 1.6235e-02, -8.6060e-03], + [ 9.6436e-03, 4.3335e-03, -3.9978e-03, ..., 3.8452e-03, + 2.1820e-03, -7.6599e-03], + ..., + [ 1.7456e-02, 1.2329e-02, -5.0354e-03, ..., 7.7515e-03, + -6.1035e-03, 5.2795e-03], + [-6.2866e-03, 7.4463e-03, -3.2349e-03, ..., 1.0437e-02, + -3.9673e-03, 4.2419e-03], + [ 3.4332e-03, -1.4832e-02, -5.6458e-03, ..., 2.1362e-02, + -1.2268e-02, 8.9722e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.after_proj.bias': tensor([ 0.0085, -0.0193, -0.0089, ..., -0.0041, -0.0183, 0.0126], + dtype=torch.bfloat16)} +[06-26 08:14:39|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:39|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 08:14:39|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 2 checkpoint. +[06-26 08:14:39|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/seg(5136ef49-6d8d-42e8-8abf-7dac722a304a) +[06-26 08:14:39|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 3 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/23057a4167b89de89a4a397fdbf3887994d115eb/general/seg/5136ef49-6d8d-42e8-8abf-7dac722a304a_ema_bf16.pt +[06-26 08:14:40|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0143, 0.0022, 0.0062], + [-0.0022, 0.0054, -0.0327, ..., -0.0080, 0.0010, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0134, -0.0026, 0.0283], + ..., + [-0.0026, 0.0125, 0.0248, ..., 0.0437, -0.0073, -0.0192], + [ 0.0277, 0.0148, 0.0054, ..., -0.0015, -0.0091, 0.0146], + [ 0.0437, 0.0366, 0.0082, ..., 0.0016, -0.0129, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0006, -0.0310, 0.0300, ..., -0.0067, -0.0043, -0.0339], + [-0.0104, -0.0066, 0.0111, ..., -0.0004, 0.0092, 0.0520], + [-0.0063, 0.0132, 0.0398, ..., -0.0101, -0.0179, -0.0064], + ..., + [-0.0034, 0.0175, -0.0096, ..., 0.0225, 0.0045, 0.0066], + [-0.0209, -0.0081, 0.0086, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0054, 0.0354, -0.0082, ..., 0.0117, 0.0356, -0.0186], + [ 0.0513, 0.0141, -0.0010, ..., 0.0376, 0.0532, 0.0026], + [ 0.0175, -0.0347, -0.0013, ..., -0.0063, 0.0057, -0.0283], + ..., + [ 0.0354, -0.0048, -0.0209, ..., -0.0153, 0.0262, -0.0113], + [ 0.0037, 0.0244, 0.0356, ..., -0.0026, 0.0170, 0.0117], + [ 0.0486, 0.0175, 0.0043, ..., 0.0016, -0.0033, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0117, ..., -0.0075, -0.0337, 0.0076], + [ 0.0491, 0.0413, 0.0251, ..., -0.0060, 0.0393, 0.0040], + [-0.0162, 0.0039, 0.0135, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0026, 0.0261, 0.0041, ..., -0.0038, 0.0347, 0.0053], + [-0.0254, 0.0068, -0.0012, ..., 0.0096, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0124, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0033, -0.0270, -0.0166, ..., 0.0117, -0.0298, 0.0532], + [-0.0016, 0.0044, -0.0129, ..., -0.0161, 0.0150, 0.0308], + [ 0.0137, 0.0084, 0.0139, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0053, -0.0630, 0.0005, ..., 0.0074, -0.0183, 0.0214], + [ 0.0034, -0.0018, 0.0258, ..., -0.0486, -0.0349, -0.0044], + [ 0.0182, -0.0074, 0.0198, ..., 0.0152, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0036, -0.0028, -0.0006, ..., 0.0035, 0.0022, -0.0117], + [-0.0154, 0.0251, 0.0432, ..., -0.0052, 0.0034, -0.0016], + [-0.0144, -0.0203, -0.0245, ..., -0.0112, -0.0325, -0.0018], + ..., + [ 0.0050, 0.0037, 0.0417, ..., 0.0281, 0.0025, 0.0030], + [-0.0038, -0.0046, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0227, -0.0364, -0.0220, ..., 0.0062, -0.0039, -0.0091]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0156, -0.0078, -0.0481, ..., 0.0449, 0.0212, 0.0127], + [ 0.0154, -0.0227, 0.0220, ..., -0.0221, -0.0020, 0.0086], + [-0.0151, -0.0095, 0.0430, ..., -0.0101, 0.0327, -0.0112], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0146], + [-0.0025, 0.0166, -0.0063, ..., 0.0002, 0.0081, -0.0078], + [-0.0112, -0.0015, 0.0123, ..., 0.0042, -0.0036, -0.0082]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0025, -0.0386, -0.0123, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0084, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0131, 0.0019, -0.0124, ..., 0.0214, 0.0105, 0.0050], + ..., + [-0.0153, -0.0172, -0.0140, ..., 0.0095, 0.0101, 0.0184], + [-0.0154, -0.0356, 0.0201, ..., -0.0047, -0.0114, 0.0332], + [-0.0396, -0.0119, -0.0092, ..., -0.0284, -0.0142, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0276, 0.0312, 0.0136, ..., -0.0195, 0.0112, -0.0142], + [ 0.0192, 0.0081, -0.0077, ..., 0.0018, -0.0135, -0.0198], + [-0.0248, 0.0100, -0.0139, ..., 0.0090, 0.0172, 0.0061], + ..., + [ 0.0466, -0.0047, 0.0003, ..., -0.0008, 0.0239, 0.0047], + [ 0.0547, -0.0194, -0.0106, ..., -0.0017, 0.0311, -0.0044], + [ 0.0035, 0.0101, -0.0098, ..., 0.0203, 0.0170, 0.0143]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0062, -0.0061, 0.0282, ..., -0.0111, -0.0038, -0.0182], + [ 0.0022, 0.0126, -0.0315, ..., -0.0457, -0.0079, 0.0105], + [-0.0009, 0.0261, 0.0245, ..., 0.0063, 0.0074, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0156, 0.0034, -0.0067], + [ 0.0076, -0.0371, 0.0356, ..., -0.0082, 0.0410, -0.0178], + [ 0.0178, -0.0146, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0078, 0.0208, 0.0276, ..., 0.0136, 0.0151, 0.0083], + [-0.0062, 0.0025, -0.0049, ..., 0.0145, 0.0005, 0.0048], + [ 0.0056, 0.0021, 0.0242, ..., -0.0120, -0.0070, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0059, ..., 0.0248, -0.0095, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., -0.0010, -0.0295, 0.0400], + [-0.0114, 0.0130, 0.0305, ..., -0.0107, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0049, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0146, -0.0043, 0.0288, ..., 0.0150, 0.0583, 0.0001], + [-0.0281, 0.0156, -0.0386, ..., -0.0444, -0.0093, -0.0013], + ..., + [ 0.0030, 0.0157, 0.0217, ..., -0.0005, 0.0115, 0.0151], + [ 0.0081, 0.0195, -0.0283, ..., -0.0009, 0.0021, -0.0342], + [-0.0162, -0.0195, -0.0013, ..., -0.0024, 0.0165, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0146, -0.0138, 0.0459, ..., -0.0018, -0.0320, 0.0278], + [ 0.0019, 0.0045, 0.0270, ..., -0.0220, -0.0183, -0.0271], + [ 0.0221, 0.0217, -0.0201, ..., 0.0030, 0.0062, -0.0053], + ..., + [-0.0258, 0.0034, 0.0215, ..., -0.0018, 0.0237, -0.0266], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0073], + [ 0.0026, -0.0082, -0.0130, ..., -0.0334, 0.0057, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0155, -0.0454, ..., -0.0134, -0.0128, 0.0400], + [ 0.0291, -0.0189, -0.0110, ..., -0.0288, 0.0025, -0.0186], + [ 0.0043, 0.0547, -0.0019, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0275, 0.0042, -0.0269, ..., -0.0096, -0.0287, 0.0150], + [ 0.0101, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0095, 0.0108, ..., -0.0058, 0.0703, -0.0052]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0091, -0.0315, ..., 0.0074, 0.0254, 0.0359], + [ 0.0079, 0.0051, 0.0040, ..., -0.0119, -0.0233, -0.0065], + [ 0.0046, -0.0037, -0.0118, ..., -0.0125, -0.0072, 0.0084], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, -0.0026], + [ 0.0339, -0.0066, 0.0223, ..., 0.0410, 0.0002, 0.0128], + [-0.0049, -0.0439, -0.0486, ..., 0.0137, -0.0129, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0053, 0.0131, -0.0023, ..., -0.0093, -0.0001, -0.0229], + [-0.0287, 0.0229, 0.0075, ..., -0.0137, 0.0045, 0.0222], + [ 0.0065, 0.0288, -0.0189, ..., -0.0173, 0.0311, -0.0262], + ..., + [ 0.0233, 0.0070, 0.0038, ..., -0.0023, 0.0278, 0.0064], + [-0.0109, 0.0005, 0.0151, ..., -0.0175, 0.0444, 0.0425], + [ 0.0062, -0.0068, 0.0093, ..., -0.0040, -0.0344, 0.0226]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[ 0.0059, 0.0125, 0.0190, ..., -0.0095, -0.0198, 0.0142], + [-0.0581, 0.0294, -0.0038, ..., -0.0240, 0.0066, 0.0302], + [ 0.0079, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0017, -0.0289, 0.0537, ..., 0.0065, -0.0056, -0.0093], + [ 0.0053, -0.0080, 0.0209, ..., -0.0035, 0.0090, 0.0186], + [-0.0082, -0.0349, -0.0173, ..., -0.0400, 0.0037, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0178, -0.0259, -0.0102, ..., 0.0090, -0.0137, -0.0405], + [ 0.0063, -0.0095, 0.0266, ..., -0.0400, 0.0047, 0.0391], + [ 0.0059, 0.0359, 0.0283, ..., 0.0295, -0.0233, 0.0212], + ..., + [ 0.0071, 0.0124, 0.0132, ..., -0.0139, -0.0033, 0.0095], + [ 0.0299, -0.0167, -0.0028, ..., -0.0052, -0.0032, -0.0206], + [ 0.0072, -0.0067, -0.0038, ..., -0.0032, -0.0047, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, 0.0056, 0.0116, ..., -0.0032, -0.0211, -0.0064], + [-0.0101, -0.0284, -0.0192, ..., 0.0376, 0.0210, 0.0229], + [ 0.0139, -0.0278, 0.0116, ..., 0.0396, -0.0061, -0.0069], + ..., + [ 0.0029, 0.0034, -0.0311, ..., 0.0344, 0.0062, -0.0125], + [-0.0236, -0.0104, -0.0061, ..., 0.0332, -0.0117, -0.0173], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0138, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[ 0.0014, 0.0166, 0.0006, ..., 0.0273, -0.0352, 0.0152], + [-0.0510, -0.0082, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0074, ..., -0.0388, 0.0004, 0.0398], + ..., + [-0.0160, -0.0195, -0.0021, ..., -0.0015, -0.0051, 0.0033], + [-0.0187, 0.0042, 0.0098, ..., 0.0287, -0.0120, -0.0101], + [-0.0181, 0.0039, 0.0247, ..., -0.0149, 0.0066, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0085, 0.0096, 0.0140, ..., -0.0112, -0.0023, -0.0253], + [-0.0166, 0.0330, -0.0171, ..., 0.0024, -0.0403, -0.0110], + [ 0.0086, 0.0347, 0.0275, ..., 0.0153, 0.0068, 0.0088], + ..., + [ 0.0271, -0.0168, -0.0105, ..., -0.0101, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0058, ..., -0.0040, 0.0006, -0.0243], + [-0.0083, 0.0581, 0.0061, ..., 0.0147, 0.0239, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0124, -0.0410, 0.0128, ..., -0.0254, -0.0033, -0.0173], + [ 0.0476, 0.0315, 0.0141, ..., -0.0148, -0.0069, 0.0049], + [-0.0096, -0.0270, 0.0089, ..., 0.0332, 0.0148, -0.0081], + ..., + [-0.0051, -0.0019, 0.0162, ..., -0.0177, 0.0171, -0.0132], + [ 0.0052, -0.0187, -0.0068, ..., 0.0248, -0.0142, -0.0076], + [-0.0143, -0.0047, 0.0150, ..., -0.0028, -0.0015, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0041, 0.0099, ..., 0.0255, -0.0253, -0.0129], + [-0.0042, -0.0139, -0.0160, ..., -0.0108, -0.0403, -0.0408], + [ 0.0175, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0113, 0.0052, 0.0322, ..., -0.0183, -0.0053, 0.0024], + [ 0.0053, -0.0057, 0.0162, ..., 0.0043, 0.0255, 0.0043], + [ 0.0270, 0.0085, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0075, ..., -0.0532, 0.0104, 0.0029], + [-0.0240, -0.0215, -0.0106, ..., -0.0050, 0.0097, -0.0168], + [ 0.0077, -0.0034, -0.0061, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0071, 0.0172, -0.0106, ..., -0.0300, 0.0005, 0.0435], + [ 0.0052, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0066], + [-0.0167, 0.0125, 0.0102, ..., -0.0247, 0.0120, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0046, -0.0081, 0.0317, ..., 0.0200, -0.0107, -0.0376], + [ 0.0356, 0.0378, 0.0135, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., 0.0027, 0.0050, -0.0322], + ..., + [ 0.0069, 0.0109, -0.0027, ..., 0.0366, -0.0206, 0.0374], + [-0.0054, -0.0080, 0.0038, ..., -0.0096, 0.0086, 0.0192], + [ 0.0099, -0.0178, -0.0038, ..., -0.0166, -0.0042, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0063, -0.0096, -0.0344, ..., -0.0049, 0.0347, -0.0181], + [ 0.0262, -0.0366, -0.0093, ..., 0.0216, 0.0349, -0.0067], + [-0.0100, -0.0286, 0.0056, ..., 0.0061, -0.0282, -0.0006], + ..., + [-0.0128, -0.0016, 0.0043, ..., -0.0089, 0.0165, 0.0076], + [-0.0322, -0.0082, -0.0170, ..., 0.0195, 0.0212, -0.0135], + [-0.0139, 0.0070, -0.0170, ..., -0.0129, 0.0059, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0043, -0.0065, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0115, 0.0374, ..., -0.0153, -0.0188, -0.0138], + ..., + [ 0.0262, 0.0003, -0.0069, ..., 0.0003, -0.0006, -0.0165], + [-0.0222, -0.0156, -0.0437, ..., -0.0065, -0.0165, -0.0282], + [-0.0130, -0.0146, 0.0029, ..., 0.0023, -0.0234, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0144, -0.0649, 0.0102], + [ 0.0249, 0.0205, -0.0050, ..., -0.0064, -0.0091, -0.0103], + [-0.0220, 0.0038, 0.0051, ..., 0.0306, 0.0381, -0.0036], + ..., + [ 0.0200, -0.0042, 0.0170, ..., 0.0118, 0.0320, -0.0157], + [ 0.0349, 0.0121, 0.0197, ..., 0.0398, 0.0315, -0.0217], + [-0.0081, 0.0256, 0.0078, ..., 0.0205, -0.0070, -0.0100]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0065, 0.0199, ..., -0.0063, 0.0276, -0.0312], + [ 0.0178, -0.0055, 0.0096, ..., 0.0113, -0.0398, 0.0293], + [ 0.0430, -0.0120, 0.0085, ..., -0.0029, -0.0067, 0.0060], + ..., + [ 0.0193, -0.0109, -0.0112, ..., -0.0049, 0.0330, 0.0143], + [-0.0122, -0.0542, -0.0205, ..., 0.0034, -0.0159, 0.0320], + [ 0.0303, -0.0039, -0.0425, ..., -0.0064, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0067, -0.0315, ..., 0.0525, -0.0052, -0.0349], + [-0.0031, -0.0032, -0.0156, ..., 0.0055, -0.0240, 0.0305], + [-0.0256, 0.0006, 0.0060, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0053, 0.0231, 0.0209, ..., 0.0073, 0.0050, 0.0461], + [-0.0078, 0.0006, 0.0060, ..., -0.0217, 0.0347, -0.0021], + [ 0.0051, 0.0234, -0.0025, ..., 0.0261, -0.0041, 0.0103]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0037, 0.0006], + [-0.0183, -0.0211, -0.0049, ..., -0.0145, 0.0146, -0.0162], + [-0.0060, -0.0160, 0.0062, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0098, 0.0064, ..., 0.0171, -0.0042, -0.0064], + [-0.0142, 0.0260, -0.0083, ..., 0.0067, 0.0104, -0.0166], + [ 0.0095, -0.0142, -0.0040, ..., 0.0062, 0.0078, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0074, 0.0143, -0.0065, ..., -0.0278, 0.0240, -0.0118], + [ 0.0012, 0.0188, -0.0160, ..., -0.0317, -0.0069, 0.0182], + [ 0.0071, 0.0310, -0.0093, ..., -0.0226, 0.0023, 0.0334], + ..., + [-0.0092, 0.0186, -0.0053, ..., 0.0228, 0.0063, 0.0050], + [ 0.0271, 0.0123, 0.0354, ..., -0.0062, 0.0198, 0.0143], + [ 0.0282, 0.0171, 0.0317, ..., -0.0113, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0110, -0.0265, -0.0115, ..., 0.0049, 0.0038, 0.0132], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0085], + [-0.0044, -0.0190, -0.0063, ..., 0.0166, -0.0135, 0.0137], + ..., + [-0.0200, -0.0009, 0.0057, ..., -0.0339, 0.0361, 0.0006], + [-0.0082, -0.0250, 0.0151, ..., -0.0035, -0.0162, 0.0220], + [-0.0075, 0.0022, 0.0205, ..., -0.0065, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0066, 0.0240, -0.0217, ..., 0.0172, 0.0045, 0.0109], + [-0.0386, 0.0103, 0.0045, ..., -0.0055, 0.0322, 0.0087], + [-0.0427, -0.0052, -0.0042, ..., 0.0062, 0.0110, -0.0135], + ..., + [-0.0085, -0.0044, -0.0264, ..., -0.0586, -0.0630, -0.0129], + [ 0.0042, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0106, -0.0610, 0.0162, ..., -0.0291, -0.0166, -0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0122, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0114], + [ 0.0112, -0.0101, -0.0327, ..., 0.0447, 0.0074, 0.0076], + [ 0.0053, -0.0175, 0.0070, ..., -0.0082, 0.0004, -0.0060], + ..., + [ 0.0295, -0.0042, -0.0103, ..., 0.0100, -0.0160, -0.0288], + [ 0.0085, 0.0253, -0.0177, ..., 0.0098, 0.0081, 0.0056], + [-0.0139, -0.0064, 0.0136, ..., -0.0128, 0.0277, 0.0086]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0061, -0.0276, -0.0271], + [ 0.0135, 0.0064, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0085, 0.0234, 0.0076, ..., 0.0464, 0.0066, 0.0309], + ..., + [-0.0123, -0.0175, 0.0327, ..., 0.0132, 0.0025, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, -0.0034, -0.0444, ..., 0.0058, -0.0283, 0.0029]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0067, -0.0104, 0.0148, ..., -0.0742, -0.0073, 0.0161], + [-0.0079, -0.0133, -0.0060, ..., 0.0056, 0.0049, 0.0239], + [ 0.0189, 0.0221, -0.0090, ..., 0.0049, 0.0408, 0.0049], + ..., + [ 0.0057, 0.0247, 0.0396, ..., 0.0057, -0.0304, -0.0183], + [ 0.0299, -0.0115, -0.0417, ..., -0.0049, -0.0376, 0.0041], + [-0.0210, 0.0168, -0.0003, ..., 0.0063, 0.0178, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[-0.0018, 0.0242, -0.0083, ..., -0.0245, 0.0064, -0.0058], + [ 0.0190, -0.0127, 0.0332, ..., -0.0146, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0079, ..., 0.0003, -0.0093, 0.0182], + ..., + [-0.0430, -0.0403, -0.0087, ..., 0.0135, -0.0141, -0.0055], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0095, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0061, 0.0122, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0122, ..., -0.0160, -0.0272, 0.0299], + [ 0.0095, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0076, ..., 0.0101, -0.0223, -0.0067], + ..., + [ 0.0192, 0.0292, -0.0026, ..., 0.0176, -0.0197, 0.0059], + [ 0.0265, 0.0063, 0.0067, ..., 0.0286, -0.0018, -0.0267], + [-0.0184, -0.0088, -0.0049, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0083, -0.0159, 0.0084, ..., 0.0198, 0.0059, 0.0277], + [ 0.0172, 0.0039, 0.0039, ..., -0.0035, 0.0021, 0.0022], + [-0.0059, 0.0184, 0.0111, ..., 0.0099, -0.0046, -0.0077], + ..., + [-0.0050, -0.0112, -0.0041, ..., -0.0096, -0.0050, 0.0118], + [-0.0427, -0.0391, -0.0192, ..., 0.0089, -0.0238, -0.0084], + [-0.0042, -0.0106, -0.0222, ..., -0.0164, 0.0342, 0.0090]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0087, -0.0161, 0.0119], + [ 0.0041, 0.0074, 0.0312, ..., 0.0074, 0.0286, 0.0317], + [ 0.0214, -0.0145, 0.0229, ..., -0.0082, 0.0056, 0.0277], + ..., + [ 0.0170, -0.0082, 0.0471, ..., 0.0072, 0.0076, -0.0046], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0067, -0.0203], + [ 0.0079, 0.0083, -0.0037, ..., -0.0089, -0.0074, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[ 0.0036, -0.0154, 0.0284, ..., -0.0104, 0.0135, -0.0042], + [ 0.0153, 0.0332, 0.0337, ..., 0.0223, -0.0165, 0.0051], + [-0.0068, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0069, 0.0077, 0.0280, ..., -0.0041, -0.0078, 0.0140], + [-0.0066, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0092, ..., 0.0064, 0.0167, -0.0086]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0079, ..., -0.0238, -0.0099, 0.0118], + [-0.0250, -0.0162, 0.0029, ..., -0.0085, -0.0020, -0.0369], + [ 0.0019, 0.0141, 0.0074, ..., -0.0452, 0.0014, 0.0119], + ..., + [ 0.0149, -0.0140, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0096, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0143, 0.0039, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[-0.0063, -0.0327, -0.0190, ..., 0.0078, 0.0013, -0.0209], + [ 0.0713, 0.0153, -0.0053, ..., -0.0035, 0.0186, 0.0002], + [-0.0227, -0.0094, 0.0188, ..., -0.0128, 0.0193, 0.0068], + ..., + [ 0.0114, 0.0464, 0.0082, ..., 0.0248, 0.0168, -0.0040], + [ 0.0381, -0.0038, 0.0063, ..., -0.0043, -0.0113, -0.0037], + [ 0.0292, 0.0044, 0.0209, ..., -0.0064, 0.0234, 0.0123]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0049, 0.0212, ..., 0.0305, 0.0136, 0.0105], + [-0.0077, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0049], + ..., + [ 0.0084, -0.0275, 0.0225, ..., 0.0055, -0.0189, -0.0105], + [-0.0125, 0.0047, -0.0058, ..., 0.0038, -0.0036, 0.0211], + [-0.0061, 0.0038, -0.0060, ..., -0.0110, 0.0264, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[ 4.6387e-03, -7.4768e-03, -1.0864e-02, ..., 1.8555e-02, + -2.1606e-02, 5.0964e-03], + [-3.8300e-03, 1.2207e-02, -2.3071e-02, ..., 1.2329e-02, + -1.1230e-02, -1.8311e-02], + [ 1.0681e-02, 1.0315e-02, 3.9041e-06, ..., 4.2480e-02, + 3.0884e-02, -2.7344e-02], + ..., + [ 3.3203e-02, 1.2390e-02, 1.4587e-02, ..., 5.3955e-02, + 4.1504e-02, 9.7046e-03], + [ 3.8605e-03, 8.5449e-03, -1.4038e-02, ..., 4.4250e-03, + -1.8921e-02, -2.3682e-02], + [-4.3030e-03, 3.5400e-02, 1.7700e-02, ..., 3.6133e-02, + -3.1128e-02, 6.9885e-03]], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0061, 0.0359, ..., -0.0090, -0.0128, -0.0022], + [ 0.0190, -0.0052, 0.0298, ..., 0.0125, -0.0043, -0.0200], + [ 0.0116, -0.0109, 0.0177, ..., -0.0248, 0.0229, 0.0014], + ..., + [-0.0082, 0.0022, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, -0.0047, -0.0291, ..., 0.0206, 0.0304, -0.0090], + [-0.0072, 0.0245, 0.0047, ..., -0.0413, 0.0216, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0062, -0.0157, -0.0065, ..., -0.0067, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0069], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, -0.0038], + ..., + [ 0.0046, -0.0049, 0.0258, ..., -0.0310, 0.0090, -0.0361], + [ 0.0061, 0.0308, 0.0542, ..., -0.0542, -0.0155, 0.0259], + [-0.0022, 0.0157, -0.0074, ..., -0.0018, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0079, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0107, ..., 0.0103, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0042, ..., 0.0123, -0.0043, 0.0089], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [ 0.0033, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0126, 0.0206, ..., 0.0043, -0.0122, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0096, -0.0231, ..., -0.0214, 0.0447, 0.0136], + [-0.0171, -0.0064, 0.0077, ..., 0.0043, 0.0114, 0.0598], + [-0.0047, 0.0267, -0.0033, ..., 0.0284, -0.0074, -0.0364], + ..., + [-0.0050, -0.0033, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0037, -0.0091, ..., 0.0211, 0.0138, -0.0076], + [ 0.0100, 0.0236, -0.0206, ..., -0.0134, 0.0046, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0064, 0.0294, ..., 0.0033, 0.0270, 0.0055], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0061, -0.0168], + [ 0.0201, -0.0121, 0.0124, ..., -0.0381, -0.0247, -0.0070], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0075], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0066, -0.0466], + [ 0.0327, 0.0284, -0.0048, ..., 0.0145, 0.0140, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0138, -0.0038, -0.0167, ..., -0.0034, 0.0170, 0.0108], + [ 0.0037, -0.0002, 0.0295, ..., -0.0172, 0.0009, 0.0135], + [ 0.0193, 0.0236, -0.0123, ..., 0.0049, -0.0369, 0.0239], + ..., + [ 0.0153, -0.0413, -0.0125, ..., -0.0354, 0.0059, -0.0320], + [-0.0266, -0.0062, -0.0327, ..., -0.0020, 0.0045, 0.0047], + [ 0.0298, 0.0356, -0.0192, ..., 0.0042, -0.0004, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0077, 0.0059, 0.0317, ..., 0.0049, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0079, -0.0096], + [-0.0166, 0.0073, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0050, 0.0260, ..., 0.0159, 0.0034, -0.0474], + [ 0.0038, 0.0070, -0.0074, ..., 0.0140, 0.0557, 0.0044], + [-0.0203, 0.0273, -0.0075, ..., -0.0330, -0.0016, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0072, ..., 0.0039, -0.0093, -0.0277], + [ 0.0057, -0.0061, -0.0058, ..., -0.0060, 0.0140, -0.0161], + [-0.0047, 0.0168, -0.0055, ..., -0.0092, 0.0124, 0.0036], + ..., + [ 0.0398, 0.0452, 0.0251, ..., -0.0067, -0.0330, -0.0012], + [-0.0620, -0.0040, 0.0371, ..., 0.0410, 0.0041, -0.0173], + [ 0.0126, -0.0039, -0.0270, ..., -0.0043, 0.0388, -0.0103]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0127, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0005, 0.0245], + [-0.0449, 0.0017, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0042, ..., 0.0041, 0.0288, 0.0127], + [ 0.0330, 0.0028, 0.0062, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0078, -0.0033, ..., -0.0060, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0066, 0.0040, 0.0527], + [ 0.0037, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0099, -0.0095, 0.0327], + [ 0.0135, 0.0113, -0.0067, ..., 0.0093, -0.0178, -0.0254], + [-0.0530, -0.0125, 0.0281, ..., -0.0242, -0.0128, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, 0.0029], + [-0.0052, -0.0025, -0.0286, ..., 0.0053, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0052, ..., 0.0018, 0.0083, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0099, 0.0150], + [ 0.0014, 0.0037, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0079, 0.0076, ..., 0.0092, -0.0111, 0.0041]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0075, -0.0038, ..., -0.0220, -0.0145, -0.0047], + [-0.0166, 0.0189, -0.0050, ..., 0.0096, 0.0016, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0043, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0070, ..., -0.0049, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0065, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0125, 0.0289, 0.0125, ..., 0.0135, 0.0065, -0.0105], + [ 0.0188, -0.0197, -0.0054, ..., 0.0284, -0.0342, -0.0066], + [-0.0525, -0.0214, 0.0153, ..., -0.0270, 0.0083, 0.0254], + ..., + [-0.0055, 0.0085, 0.0281, ..., 0.0129, -0.0035, -0.0065], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0046, 0.0033], + [-0.0273, -0.0747, 0.0208, ..., -0.0047, 0.0046, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0056, 0.0148, 0.0334, ..., 0.0205, -0.0273, -0.0020], + [-0.0165, 0.0110, -0.0135, ..., 0.0106, -0.0067, 0.0219], + [ 0.0046, 0.0227, -0.0055, ..., 0.0126, -0.0061, 0.0151], + ..., + [-0.0347, -0.0148, 0.0062, ..., 0.0040, 0.0030, -0.0093], + [-0.0073, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0039, -0.0500, -0.0232, ..., 0.0320, 0.0037, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0052, -0.0364], + [ 0.0220, -0.0026, -0.0115, ..., -0.0043, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0117, ..., 0.0444, 0.0049, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0067, 0.0192, -0.0393], + [-0.0102, -0.0244, 0.0275, ..., 0.0038, 0.0043, 0.0081], + [-0.0170, -0.0427, 0.0132, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0050, -0.0378, -0.0359, ..., 0.0552, -0.0075, -0.0160], + [ 0.0047, 0.0142, 0.0134, ..., 0.0055, -0.0153, 0.0354], + [ 0.0094, 0.0040, 0.0057, ..., 0.0028, -0.0249, -0.0136], + ..., + [ 0.0068, -0.0125, 0.0049, ..., -0.0095, -0.0079, -0.0036], + [-0.0073, 0.0439, 0.0166, ..., 0.0248, 0.0134, -0.0187], + [ 0.0175, 0.0061, 0.0049, ..., 0.0410, -0.0096, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0107, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0142, 0.0038, ..., -0.0035, -0.0244, 0.0356], + [ 0.0071, 0.0165, -0.0200, ..., 0.0033, 0.0237, -0.0078], + ..., + [-0.0197, 0.0054, 0.0386, ..., 0.0181, 0.0109, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, -0.0047, -0.0374], + [-0.0085, -0.0095, 0.0172, ..., 0.0243, -0.0219, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-3.4668e-02, 6.2866e-03, 1.9287e-02, ..., -2.2095e-02, + -1.4771e-02, -2.1118e-02], + [ 1.3062e-02, 1.1658e-02, 2.5269e-02, ..., 2.0386e-02, + -2.3193e-02, -3.2227e-02], + [ 2.5513e-02, -4.8523e-03, 4.1809e-03, ..., -5.7068e-03, + 5.4016e-03, 3.3691e-02], + ..., + [ 4.2236e-02, 2.4414e-03, 3.5156e-02, ..., -3.0396e-02, + -4.7207e-05, -2.0386e-02], + [ 3.0151e-02, 1.5503e-02, 9.9487e-03, ..., -7.2327e-03, + -2.5391e-02, -2.4536e-02], + [-6.4087e-04, -6.2561e-03, -2.0599e-03, ..., 4.1504e-03, + -1.9653e-02, -2.8198e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0042, 0.0266, ..., 0.0155, 0.0187, 0.0047], + [-0.0118, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0036, ..., -0.0172, -0.0092, 0.0231], + ..., + [-0.0295, -0.0136, -0.0036, ..., 0.0042, -0.0048, 0.0101], + [-0.0048, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0078], + [-0.0015, -0.0043, -0.0259, ..., -0.0030, -0.0141, -0.0115]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0040, 0.0064, -0.0140, ..., 0.0242, -0.0040, -0.0386], + [-0.0142, 0.0068, -0.0101, ..., -0.0369, -0.0151, 0.0157], + [-0.0267, -0.0045, -0.0074, ..., -0.0126, 0.0182, -0.0115], + ..., + [ 0.0172, -0.0038, 0.0144, ..., 0.0115, -0.0024, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., -0.0023, -0.0033, -0.0137], + [-0.0219, -0.0396, 0.0039, ..., -0.0054, -0.0037, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-3.3417e-03, 1.6846e-02, 1.9409e-02, ..., 3.4912e-02, + -4.1992e-02, 2.0142e-03], + [ 4.3945e-03, 2.7344e-02, 1.4526e-02, ..., -1.8433e-02, + -3.5248e-03, 1.5442e-02], + [ 1.4465e-02, -3.7231e-03, -7.3242e-03, ..., 2.4536e-02, + 3.5889e-02, 5.8289e-03], + ..., + [ 4.3213e-02, 1.4526e-02, -3.1982e-02, ..., 8.8215e-05, + 4.2114e-03, 2.6611e-02], + [ 2.2949e-02, 1.2573e-02, -1.9897e-02, ..., -7.5989e-03, + -3.1250e-02, -4.5654e-02], + [ 4.9133e-03, 1.0910e-03, -6.8359e-03, ..., 4.9133e-03, + -1.0437e-02, -1.5503e-02]], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0115, 0.0388, 0.0359], + [-0.0042, 0.0121, -0.0019, ..., 0.0123, -0.0046, 0.0068], + [-0.0277, 0.0011, -0.0181, ..., -0.0003, -0.0036, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0035, 0.0513, -0.0109], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0110], + [-0.0282, -0.0046, 0.0119, ..., -0.0055, -0.0038, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0033, -0.0128, ..., 0.0140, 0.0308, 0.0059], + [ 0.0120, -0.0332, -0.0320, ..., -0.0075, -0.0140, -0.0032], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, 0.0029, -0.0311, ..., 0.0124, 0.0059, 0.0081], + [ 0.0040, 0.0198, 0.0141, ..., -0.0222, -0.0143, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, 0.0049, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0090, -0.0035, -0.0027, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0049, ..., 0.0302, -0.0018, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0023, -0.0036]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0028, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0037, -0.0222, 0.0262], + ..., + [-0.0147, 0.0043, 0.0260, ..., -0.0042, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0066, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0035, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0112], + [-0.0278, 0.0041, -0.0374, ..., 0.0276, -0.0003, 0.0151], + [ 0.0089, 0.0237, -0.0039, ..., 0.0527, -0.0031, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0072, -0.0095], + [-0.0067, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0036, -0.0128, ..., 0.0063, -0.0371, -0.0064], + [ 0.0064, -0.0130, -0.0028, ..., -0.0204, -0.0013, 0.0057], + [-0.0112, -0.0195, -0.0053, ..., -0.0542, -0.0094, -0.0059], + ..., + [-0.0205, -0.0082, -0.0037, ..., -0.0199, 0.0044, 0.0058], + [-0.0220, -0.0302, -0.0105, ..., 0.0114, 0.0369, 0.0052], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0134, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[ 0.0017, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0034, 0.0056, 0.0238, ..., -0.0140, -0.0320, 0.0041], + [-0.0193, 0.0364, -0.0035, ..., -0.0076, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0065, -0.0232, -0.0464], + [-0.0107, -0.0298, -0.0195, ..., 0.0109, 0.0312, 0.0151], + [-0.0074, -0.0116, -0.0195, ..., 0.0034, 0.0107, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0149, ..., -0.0371, 0.0400, -0.0165], + [ 0.0047, -0.0239, -0.0123, ..., -0.0062, 0.0243, -0.0396], + [-0.0101, 0.0155, -0.0038, ..., -0.0277, 0.0042, 0.0037], + ..., + [ 0.0112, -0.0047, 0.0515, ..., 0.0327, -0.0120, -0.0077], + [-0.0030, 0.0170, 0.0063, ..., -0.0014, 0.0280, -0.0210], + [-0.0016, 0.0076, 0.0055, ..., 0.0153, -0.0044, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0063, 0.0081, -0.0077, ..., -0.0588, 0.0187, -0.0068], + [-0.0102, 0.0199, -0.0077, ..., 0.0080, -0.0332, -0.0026], + ..., + [-0.0051, -0.0464, 0.0277, ..., -0.0143, 0.0198, 0.0071], + [ 0.0162, -0.0053, 0.0082, ..., -0.0105, -0.0038, 0.0297], + [-0.0269, 0.0129, 0.0164, ..., 0.0086, 0.0176, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0033, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, 0.0028, ..., 0.0142, 0.0128, 0.0439], + [-0.0352, 0.0042, -0.0201, ..., 0.0330, 0.0309, -0.0049], + ..., + [-0.0033, 0.0123, 0.0153, ..., -0.0034, -0.0038, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0039, 0.0175], + [-0.0244, 0.0036, -0.0300, ..., 0.0051, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0046, -0.0019, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0029, ..., 0.0310, 0.0088, -0.0023], + [ 0.0033, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0109, -0.0237, ..., -0.0104, 0.0157, -0.0037], + [-0.0081, 0.0205, 0.0140, ..., 0.0262, -0.0195, 0.0077], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, 0.0040, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0018, 0.0034, -0.0093], + [-0.0009, 0.0019, 0.0398, ..., -0.0304, -0.0029, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0019], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0042, 0.0065, 0.0228, ..., -0.0322, 0.0315, -0.0012], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0050, 0.0291], + [ 0.0135, -0.0014, -0.0339, ..., 0.0074, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, -0.0024], + [-0.0166, -0.0043, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0065, -0.0039, ..., 0.0240, 0.0037, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, -0.0020], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0014, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0011, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0066, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0022], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, -0.0039, 0.0036], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, 0.0027, ..., -0.0164, -0.0020, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0033, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0027, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0056, -0.0231, -0.0255, ..., 0.0038, 0.0295, -0.0057], + [-0.0255, 0.0513, 0.0029, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0031, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0067, -0.0070, -0.0364, ..., -0.0175, 0.0306, -0.0078], + [-0.0051, -0.0195, -0.0012, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0075, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0072, -0.0193, -0.0140], + [-0.0166, 0.0062, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0133, -0.0056, 0.0145, ..., -0.0072, 0.0211, -0.0275], + [ 0.0078, -0.0057, -0.0234, ..., 0.0084, 0.0071, -0.0166], + [ 0.0225, -0.0251, 0.0052, ..., 0.0078, 0.0236, -0.0046], + ..., + [ 0.0186, 0.0143, -0.0046, ..., 0.0272, -0.0083, -0.0096], + [ 0.0077, 0.0035, 0.0072, ..., -0.0260, 0.0267, 0.0151], + [ 0.0103, -0.0076, 0.0212, ..., 0.0195, -0.0157, 0.0121]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0087, -0.0270, 0.0251, ..., 0.0096, 0.0049, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.3.proj.1.weight': tensor([[-0.0172, 0.0071, 0.0757, ..., -0.0610, 0.0327, -0.0320], + [-0.0278, 0.0029, -0.1128, ..., 0.0272, -0.0361, 0.0583], + [-0.0347, 0.0947, -0.0361, ..., -0.0166, 0.0172, -0.0332], + ..., + [ 0.0125, 0.0007, -0.0100, ..., 0.0219, -0.0376, -0.0036], + [-0.0239, -0.0574, 0.0586, ..., 0.0378, 0.0830, 0.0608], + [ 0.0176, -0.0205, 0.0076, ..., 0.0248, 0.0493, 0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_proj.weight': tensor([[-4.1199e-03, 5.4932e-03, 6.3782e-03, ..., 2.3346e-03, + -5.4359e-05, 8.1177e-03], + [-9.6893e-04, 8.4305e-04, -8.3923e-04, ..., -8.9111e-03, + -4.5586e-04, 2.2736e-03], + [ 4.7913e-03, -3.5553e-03, 3.2806e-03, ..., -8.0566e-03, + -4.3335e-03, 1.3809e-03], + ..., + [-2.3193e-03, -4.1504e-03, -7.0496e-03, ..., 1.2268e-02, + 4.0588e-03, -8.0566e-03], + [-9.6436e-03, -7.9346e-03, -1.2634e-02, ..., -1.3123e-02, + -5.9204e-03, -6.7139e-04], + [-1.7929e-03, -8.9111e-03, 6.1646e-03, ..., 9.9487e-03, + -4.2915e-04, 1.8005e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_norm.weight': tensor([ 4.7852e-02, -1.6968e-02, 4.7119e-02, 7.8613e-02, -1.3809e-03, + 4.0513e-08, -1.1362e-16, -1.8396e-21, -3.8598e-17, -1.1689e-19, + -5.3204e-21, 3.8257e-24, -6.4851e-22, -1.0727e-24, -2.7504e-23, + 3.6028e-25, -6.9384e-08, -4.4250e-03, -2.9395e-09, -8.8476e-09, + 7.6834e-09, 1.5924e-19, 2.1973e-02, 3.1281e-04, -5.5908e-02, + 4.2578e-01, -1.0889e-01, -3.9062e-03, 1.7822e-02, 8.7891e-02, + 1.0029e-18, 2.1537e-08, 2.6855e-03, -5.1737e-05, -5.5552e-05, + 5.1880e-04, -4.3074e-09, 8.7486e-14, -4.0118e-23, 2.0195e-26, + -7.4216e-27, 8.4703e-22, 3.2558e-21, -1.4343e-03, 1.9775e-02, + -3.8719e-04, -5.4688e-02, -5.0293e-02, 1.7969e-01, 2.9541e-02, + -5.5542e-03, 8.2397e-04, -3.9062e-03, 3.2596e-07, 7.0690e-17, + -7.6762e-22, 5.8981e-17, 1.3897e-22, -2.2334e-22, 9.1382e-27, + 2.7297e-22, -1.3944e-13, 1.9690e-27, 6.0269e-28, 1.2988e-01, + 1.4746e-01, 3.4570e-01, 7.7820e-03, 1.8311e-03, 2.5728e-08, + 7.2858e-17, 3.7481e-20, -1.4162e-18, -9.3512e-19, -7.4115e-20, + -1.3329e-25, 2.2658e-20, -5.5008e-23, 2.5312e-22, 7.6256e-25, + -1.3318e-07, -7.6771e-05, -2.6375e-06, 3.3424e-11, 2.0210e-07, + -1.2559e-15, -2.5635e-03, -1.6022e-03, 3.4570e-01, -1.5625e-02, + 4.1992e-01, 7.9102e-02, -8.3618e-03, 5.3711e-03, -1.2536e-18, + 1.3132e-07, 6.4087e-04, -9.7156e-06, -3.7193e-05, 5.3406e-05, + -7.4971e-08, 1.2546e-14, -2.7303e-25, -2.2877e-24, -4.2662e-27, + -1.5096e-23, 7.9939e-21, 1.2756e-02, 2.8076e-03, -1.2054e-03, + 3.9844e-01, 2.0703e-01, 1.5430e-01, -1.6113e-01, 7.7515e-03, + -4.9805e-02, -3.1128e-03, 1.5087e-07, 2.9707e-17, -5.7439e-21, + 3.4911e-17, 3.6528e-21, 7.3268e-20, 4.3823e-26, 1.4540e-24, + 1.9895e-11, 7.8065e-24, -1.0057e-25], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.self_attn.k_proj.weight': tensor([[-7.0496e-03, 4.8218e-03, -7.4768e-03, ..., -4.2114e-03, + -3.3264e-03, 2.1515e-03], + [-1.9287e-02, -1.4496e-03, 1.9226e-03, ..., 1.4114e-03, + -5.8289e-03, -1.9531e-03], + [-2.1219e-05, -1.2436e-03, 6.7139e-03, ..., 1.1826e-03, + 1.5488e-03, 8.5449e-03], + ..., + [ 2.5330e-03, 2.9602e-03, -2.6550e-03, ..., -5.2261e-04, + -1.1215e-03, 3.8910e-03], + [ 1.1368e-03, 1.0757e-03, -7.6599e-03, ..., -1.0529e-03, + -1.5354e-04, -1.1063e-03], + [ 1.4648e-03, 1.9684e-03, -3.1738e-03, ..., -5.8594e-03, + -2.7466e-04, 1.3962e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.k_norm.weight': tensor([ 9.4727e-02, 5.6641e-02, -5.5664e-02, 7.6660e-02, 1.5335e-03, + -2.1770e-08, -1.1189e-16, 1.7999e-20, -3.8381e-17, -1.6178e-19, + -8.7350e-21, -3.7740e-24, -2.6073e-21, 1.4605e-24, 1.8301e-23, + 6.4300e-25, -6.6590e-08, 4.1504e-03, 1.0477e-09, 8.8476e-09, + 7.5088e-09, -1.1181e-19, 1.6724e-02, 1.3504e-03, -1.5625e-02, + 4.2383e-01, -1.1816e-01, 7.5684e-02, 3.2349e-03, 4.5538e-05, + -1.4704e-18, 6.2864e-08, -2.6093e-03, -5.1498e-05, -5.5313e-05, + -5.4932e-04, -8.7894e-09, -7.1942e-14, 4.0325e-23, 2.8071e-26, + -6.6643e-27, -8.4372e-22, -5.1351e-21, 1.0193e-02, 1.8311e-02, + 2.2125e-04, 8.5449e-02, 1.2598e-01, 2.0801e-01, 1.6211e-01, + 7.1411e-03, 4.9072e-02, -3.7994e-03, -3.8557e-07, -4.6079e-18, + -1.6874e-21, 5.2313e-18, 4.4668e-22, -9.2644e-22, -1.1763e-26, + 2.6635e-22, -2.5757e-13, -2.0397e-26, -9.4663e-28, 8.1177e-03, + 1.4355e-01, 3.4766e-01, -4.5898e-02, 1.6556e-03, 4.7730e-08, + 5.6379e-17, 3.5999e-20, 1.6263e-18, 9.1480e-19, -7.3692e-20, + -6.8259e-26, 2.0964e-20, 5.5421e-23, -2.6304e-22, 7.8194e-25, + -1.6857e-07, -8.5354e-05, -2.6226e-06, 7.9012e-12, 1.9558e-07, + 1.2976e-15, -1.5564e-03, -3.9291e-04, 3.4570e-01, -1.6479e-02, + 4.1992e-01, 5.7983e-03, 1.7578e-02, 8.7402e-02, -1.7872e-19, + 6.9849e-08, 4.4632e-04, 1.0669e-05, -3.7432e-05, 4.8637e-05, + -7.9628e-08, -1.2379e-14, -1.7933e-25, -2.0550e-24, 4.2914e-27, + 1.5303e-23, -7.9409e-21, 8.2397e-03, 1.1292e-02, 7.4387e-04, + 4.0039e-01, 1.8359e-01, 1.4355e-01, 1.5625e-02, 7.4768e-03, + -1.1597e-03, -3.6621e-03, -8.1491e-08, -3.8164e-17, 5.7175e-21, + -3.7513e-17, -3.4940e-21, -7.1998e-20, 2.3426e-26, -2.2101e-24, + 1.9895e-11, 7.8065e-24, -1.0138e-25], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.self_attn.v_proj.weight': tensor([[-0.0040, 0.0033, -0.0051, ..., -0.0042, 0.0036, -0.0044], + [ 0.0073, -0.0018, 0.0019, ..., -0.0043, -0.0013, -0.0023], + [ 0.0152, 0.0034, -0.0048, ..., 0.0084, -0.0022, -0.0011], + ..., + [ 0.0050, 0.0029, -0.0007, ..., -0.0044, 0.0009, 0.0006], + [ 0.0009, 0.0044, -0.0021, ..., -0.0018, 0.0045, -0.0052], + [ 0.0008, -0.0037, 0.0006, ..., -0.0032, -0.0033, -0.0094]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.output_proj.weight': tensor([[ 0.0129, -0.0291, -0.0254, ..., 0.0038, 0.0062, 0.0021], + [ 0.0141, -0.0053, 0.0117, ..., 0.0043, 0.0077, 0.0028], + [-0.0046, 0.0078, 0.0023, ..., 0.0040, 0.0029, -0.0049], + ..., + [ 0.0094, -0.0002, -0.0016, ..., -0.0038, 0.0046, -0.0112], + [-0.0047, 0.0088, 0.0038, ..., -0.0010, 0.0015, 0.0010], + [-0.0056, 0.0040, -0.0090, ..., 0.0036, 0.0044, -0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_proj.weight': tensor([[-0.0025, -0.0159, 0.0013, ..., 0.0058, -0.0039, 0.0007], + [ 0.0072, 0.0210, 0.0029, ..., -0.0058, 0.0050, -0.0034], + [ 0.0085, 0.0194, 0.0049, ..., -0.0053, 0.0130, -0.0042], + ..., + [ 0.0072, 0.0052, 0.0026, ..., 0.0013, -0.0083, 0.0012], + [-0.0029, -0.0007, -0.0029, ..., -0.0027, 0.0109, -0.0044], + [ 0.0002, -0.0038, 0.0008, ..., -0.0022, 0.0015, 0.0057]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_norm.weight': tensor([ 2.0695e-04, 1.8677e-02, 2.3192e-11, 5.6458e-04, 3.6049e-04, + -9.7156e-06, -6.0499e-06, 2.2125e-03, 3.6377e-02, 2.6611e-02, + 1.4305e-04, 3.8330e-02, 5.5859e-01, 3.2196e-03, 8.4839e-03, + 1.1981e-05, 4.2969e-01, 6.8188e-05, 1.9360e-04, 1.1292e-02, + 3.8528e-04, 1.5640e-03, 9.1309e-02, 1.1902e-02, 4.0283e-03, + 6.3171e-03, 2.3682e-02, 3.0696e-06, 9.1553e-03, 2.3651e-04, + 3.6133e-01, 1.9165e-02, 6.0120e-03, 3.8281e-01, 5.3125e-01, + 9.4986e-04, 8.8882e-04, 2.5879e-02, 1.4258e-01, 8.6060e-03, + 1.6113e-02, 2.0142e-03, -4.4890e-07, 3.3875e-03, 1.8921e-03, + 4.7119e-02, 9.3994e-03, 4.0770e-05, 2.3041e-03, 2.0218e-04, + 1.9932e-04, 1.8552e-06, 1.4156e-06, -1.7762e-05, -3.5400e-03, + 1.3672e-02, 2.6367e-02, 1.3275e-03, 1.4954e-02, 1.4526e-02, + -7.0572e-05, -2.9057e-07, 2.9945e-04, 8.1253e-04, 1.1658e-02, + 4.3640e-03, 1.5640e-03, 2.2070e-01, 1.0803e-02, 2.4780e-02, + 3.9258e-01, 1.5259e-05, 4.6143e-02, 2.0020e-02, 3.7842e-03, + -1.9550e-04, 4.4556e-03, 6.7139e-03, 6.1035e-03, 9.2697e-04, + 9.8267e-03, 1.0071e-02, 1.6016e-01, 1.7929e-03, 2.3127e-05, + 1.1719e-02, -1.5945e-03, 1.1265e-05, 8.9111e-03, 4.7852e-01, + 5.0049e-03, 6.5231e-04, 1.4282e-02, 6.4697e-03, 2.5630e-06, + 4.1406e-01, 1.6928e-05, 1.7929e-04, -1.6964e-13, 6.5613e-03, + 3.2617e-01, -2.6524e-06, -4.1199e-03, 4.0283e-03, 3.4912e-02, + 6.1512e-05, -2.4199e-05, 1.8477e-05, 1.9775e-02, 1.0132e-02, + 1.3504e-03, -3.4332e-05, 9.3994e-03, 4.8218e-03, 2.2430e-03, + 7.1335e-04, 2.0874e-02, 1.7471e-03, -9.5367e-04, 5.3101e-03, + 4.9353e-05, 4.5471e-03, 9.1934e-04, 1.2054e-03, 1.5503e-02, + 1.4210e-04, 2.7161e-03, 1.8501e-04], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_proj.weight': tensor([[-0.0022, 0.0007, -0.0010, ..., -0.0041, 0.0014, -0.0015], + [-0.0013, -0.0021, -0.0042, ..., -0.0038, -0.0054, -0.0001], + [ 0.0043, -0.0015, -0.0052, ..., -0.0038, -0.0040, -0.0036], + ..., + [ 0.0038, -0.0025, 0.0025, ..., 0.0010, 0.0006, -0.0012], + [ 0.0008, 0.0005, -0.0023, ..., -0.0032, 0.0006, 0.0013], + [ 0.0017, 0.0019, -0.0031, ..., -0.0045, -0.0006, 0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_norm.weight': tensor([ 1.6499e-04, 1.8677e-02, -2.0009e-11, 2.2125e-04, 3.4523e-04, + 1.2815e-05, 6.0499e-06, 2.1667e-03, 3.5645e-02, 2.6611e-02, + 1.5354e-04, 3.8574e-02, 5.6250e-01, 3.1738e-03, 8.0566e-03, + 9.1195e-06, 4.2969e-01, 6.8188e-05, -1.1015e-04, 1.1230e-02, + 6.5994e-04, 1.5945e-03, 9.1309e-02, 1.2024e-02, 3.8757e-03, + 6.3171e-03, 2.3438e-02, 2.9057e-06, 8.5449e-03, 2.3651e-04, + 3.6133e-01, 1.8921e-02, 5.9509e-03, 3.8281e-01, 5.3125e-01, + 8.7738e-04, 1.0223e-03, 2.6123e-02, 1.4355e-01, 8.7280e-03, + 1.6113e-02, 2.0142e-03, -4.8429e-07, 3.1891e-03, 1.3733e-03, + 4.6875e-02, 9.3384e-03, -1.4901e-05, 2.3193e-03, -7.9155e-05, + 2.0218e-04, 1.2815e-06, 1.5944e-06, 2.4915e-05, 3.9673e-03, + 1.3550e-02, 2.6245e-02, 1.3504e-03, 1.5259e-02, 1.4587e-02, + -7.7248e-05, -5.5507e-07, -3.1471e-04, 7.4005e-04, 1.1719e-02, + 4.3335e-03, 1.5640e-03, 2.2070e-01, 1.0742e-02, 2.4902e-02, + 3.9258e-01, 1.5259e-05, 4.4434e-02, 2.0020e-02, 3.7994e-03, + 1.9455e-04, 3.6926e-03, 6.6833e-03, 6.1646e-03, 9.1171e-04, + 9.7656e-03, 1.0071e-02, 1.6016e-01, 1.5564e-03, -1.7524e-05, + 1.1719e-02, 1.9379e-03, 1.1861e-05, 8.9722e-03, 4.7852e-01, + 4.9744e-03, 6.1798e-04, 1.4099e-02, 6.4697e-03, 2.2501e-06, + 4.1406e-01, 4.2021e-06, 7.7724e-05, 2.2116e-13, 6.5613e-03, + 3.2812e-01, 2.6673e-06, 4.2114e-03, 4.0283e-03, 3.4912e-02, + 6.2943e-05, 3.9101e-05, 1.8835e-05, 1.9165e-02, 1.0193e-02, + 1.5945e-03, 3.5048e-05, 9.3384e-03, 4.7607e-03, 2.6093e-03, + -6.7139e-04, 2.0874e-02, 1.8539e-03, 1.2131e-03, 5.3101e-03, + 3.6955e-05, 4.6997e-03, 9.1171e-04, 1.1749e-03, 1.5076e-02, + -1.4114e-04, 2.8076e-03, -5.2452e-05], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.v_proj.weight': tensor([[-7.5989e-03, -1.1215e-03, 1.2329e-02, ..., -7.8964e-04, + 5.2490e-03, -1.0742e-02], + [-2.1729e-02, -2.2278e-03, 5.7068e-03, ..., -3.0060e-03, + -8.9111e-03, -2.3956e-03], + [-9.3994e-03, -1.8768e-03, 4.5776e-03, ..., -4.2114e-03, + 1.4771e-02, -1.2085e-02], + ..., + [ 1.0010e-02, -5.7220e-04, 4.4346e-05, ..., -8.2779e-04, + 3.7384e-03, 1.9897e-02], + [-7.6294e-03, 1.6708e-03, 1.0803e-02, ..., 1.3046e-03, + -1.7334e-02, -1.2360e-03], + [-1.6846e-02, -1.8082e-03, 8.1177e-03, ..., -1.2665e-03, + -7.4768e-03, -2.2583e-02]], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.output_proj.weight': tensor([[-0.0013, -0.0039, 0.0039, ..., 0.0077, 0.0042, 0.0098], + [ 0.0042, 0.0090, -0.0073, ..., 0.0137, 0.0020, -0.0074], + [ 0.0077, -0.0078, -0.0010, ..., 0.0038, -0.0059, -0.0006], + ..., + [-0.0054, -0.0056, -0.0043, ..., -0.0075, 0.0052, -0.0055], + [ 0.0081, 0.0005, 0.0091, ..., -0.0078, -0.0079, -0.0040], + [ 0.0175, -0.0094, -0.0039, ..., -0.0023, -0.0044, 0.0064]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_img.weight': tensor([[ 0.0078, 0.0121, 0.0018, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0074, 0.0249, 0.0291], + [-0.0095, 0.0080, -0.0121, ..., -0.0342, -0.0327, 0.0065], + ..., + [ 0.0051, -0.0067, 0.0038, ..., 0.0084, 0.0286, -0.0165], + [ 0.0078, -0.0067, -0.0101, ..., -0.0095, 0.0140, -0.0309], + [ 0.0242, -0.0101, 0.0255, ..., -0.0227, 0.0075, -0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, 0.0018], + [ 0.0041, 0.0120, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0080, ..., -0.0056, -0.0057, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, 0.0045, 0.0052], + [-0.0160, 0.0042, 0.0036, ..., 0.0133, -0.0173, 0.0024], + [ 0.0038, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0054, -0.0364, ..., -0.0225, -0.0145, -0.0052], + [-0.0157, -0.0052, -0.0364, ..., 0.0095, 0.0111, -0.0023], + [ 0.0037, 0.0615, 0.0018, ..., -0.0205, -0.0306, -0.0115], + ..., + [-0.0043, 0.0042, -0.0099, ..., -0.0211, -0.0044, 0.0288], + [ 0.0266, -0.0117, 0.0047, ..., -0.0420, -0.0282, -0.0014], + [ 0.0286, 0.0325, 0.0094, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.mlp.layer1.weight': tensor([[ 0.0036, -0.0023, -0.0013, ..., -0.0089, 0.0038, 0.0076], + [-0.0030, 0.0058, -0.0032, ..., -0.0057, 0.0044, 0.0010], + [-0.0087, 0.0012, -0.0061, ..., -0.0046, 0.0040, 0.0060], + ..., + [-0.0016, -0.0054, 0.0029, ..., -0.0075, -0.0067, 0.0027], + [-0.0028, 0.0018, -0.0085, ..., 0.0128, 0.0232, 0.0038], + [ 0.0034, 0.0015, -0.0031, ..., 0.0009, -0.0068, 0.0004]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.mlp.layer2.weight': tensor([[-2.4109e-03, 7.9346e-03, -9.7656e-03, ..., -1.1215e-03, + 2.0599e-03, 1.2268e-02], + [-1.8005e-03, 6.0425e-03, 3.6011e-03, ..., -1.6846e-02, + -2.3071e-02, 4.7607e-03], + [-7.0035e-06, -8.9722e-03, -4.5166e-03, ..., -5.1575e-03, + -7.6294e-03, -1.2573e-02], + ..., + [ 2.1515e-03, 1.1902e-03, -9.5215e-03, ..., 2.8381e-03, + 8.3008e-03, -7.6904e-03], + [-1.0437e-02, 5.0354e-03, 6.3324e-04, ..., 1.0742e-02, + 3.3447e-02, 5.6152e-03], + [-3.8300e-03, 1.1230e-02, 5.2490e-03, ..., 7.4768e-03, + 5.3406e-03, 2.8610e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0476, -0.0464, -0.0449, ..., 0.0027, 0.0028, 0.0029], + [ 0.0457, 0.0447, 0.0435, ..., -0.0021, -0.0022, -0.0023], + [ 0.0454, 0.0449, 0.0439, ..., -0.0023, -0.0024, -0.0024], + ..., + [ 0.0479, 0.0471, 0.0454, ..., -0.0021, -0.0022, -0.0022], + [ 0.0471, 0.0459, 0.0444, ..., -0.0020, -0.0020, -0.0019], + [ 0.0471, 0.0461, 0.0449, ..., -0.0027, -0.0029, -0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0021, 0.0021, 0.0022, ..., 0.0049, 0.0047, 0.0020], + [-0.0017, 0.0016, 0.0016, ..., 0.0049, 0.0052, 0.0017], + [-0.0059, 0.0059, 0.0054, ..., 0.0074, 0.0082, 0.0057], + ..., + [-0.0452, 0.0525, 0.0515, ..., 0.0515, 0.0488, 0.0452], + [ 0.0047, 0.0081, 0.0080, ..., -0.0068, -0.0065, -0.0046], + [ 0.0081, -0.0093, -0.0092, ..., -0.0067, -0.0059, -0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5847e-03, 5.9509e-03, 7.9346e-03, ..., -6.8665e-04, + -6.1417e-04, -7.8201e-04], + [-3.8086e-02, -3.6621e-02, -3.5889e-02, ..., 3.4332e-04, + 3.5286e-04, 3.6049e-04], + [ 4.2725e-02, 4.1016e-02, 4.0527e-02, ..., -1.5259e-05, + -5.3346e-06, -2.0266e-05], + ..., + [ 3.7842e-02, 3.7109e-02, 3.6377e-02, ..., -4.2152e-04, + -3.5667e-04, -3.9291e-04], + [ 4.9561e-02, 4.8096e-02, 4.7363e-02, ..., -7.9632e-05, + -4.7386e-06, -6.6757e-05], + [ 3.8574e-02, 3.7354e-02, 3.6865e-02, ..., 6.9141e-05, + 1.4782e-04, 9.6321e-05]], dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0070, -0.0059, 0.0048, ..., 0.0062, 0.0064, 0.0053], + [-0.0058, -0.0038, -0.0053, ..., 0.0038, 0.0035, 0.0025], + [-0.0043, -0.0044, -0.0085, ..., 0.0043, -0.0074, -0.0071], + ..., + [-0.0078, -0.0134, 0.0259, ..., 0.0132, 0.0273, 0.0216], + [ 0.0114, 0.0058, 0.0103, ..., -0.0056, 0.0052, 0.0107], + [-0.0127, 0.0037, -0.0052, ..., -0.0037, -0.0037, -0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_mlp.1.weight': tensor([[-0.0050, -0.0051, -0.0050, ..., 0.0005, 0.0007, 0.0007], + [ 0.0530, 0.0510, 0.0498, ..., -0.0020, -0.0024, -0.0024], + [ 0.0444, 0.0432, 0.0422, ..., -0.0033, -0.0036, -0.0036], + ..., + [ 0.0894, 0.0869, 0.0859, ..., -0.0005, -0.0007, -0.0007], + [ 0.0050, 0.0049, 0.0050, ..., -0.0005, -0.0006, -0.0005], + [ 0.0640, 0.0625, 0.0605, ..., -0.0010, -0.0010, -0.0010]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_mlp.2.weight': tensor([[-0.0010, 0.0074, 0.0080, ..., 0.0094, 0.0010, 0.0070], + [ 0.0007, 0.0026, 0.0031, ..., 0.0031, -0.0007, 0.0008], + [ 0.0004, 0.0029, 0.0034, ..., 0.0034, -0.0004, 0.0034], + ..., + [ 0.0072, 0.0603, 0.0820, ..., 0.0562, -0.0073, 0.0408], + [ 0.0059, 0.0070, 0.0063, ..., 0.0101, -0.0056, 0.0061], + [-0.0122, -0.0049, -0.0046, ..., -0.0051, 0.0120, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.before_proj.weight': tensor([[ 1.0431e-05, 1.1215e-03, 9.0027e-04, ..., -2.1820e-03, + 2.7657e-04, 3.4027e-03], + [ 3.8147e-03, 5.9509e-04, 1.0681e-02, ..., 1.2207e-02, + -7.5073e-03, -2.5635e-03], + [-1.1215e-03, -3.4637e-03, 2.3346e-03, ..., 1.5030e-03, + -9.9945e-04, -7.0190e-04], + ..., + [ 5.7373e-03, 2.3193e-03, -4.4060e-04, ..., -1.9684e-03, + -1.2207e-04, -5.2490e-03], + [ 2.5635e-03, -4.5166e-03, -4.9133e-03, ..., -1.8692e-03, + 4.3030e-03, -1.2512e-03], + [ 4.6387e-03, 3.4142e-04, -9.4604e-04, ..., 2.0905e-03, + -4.0588e-03, -5.6839e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.0.before_proj.bias': tensor([-0.0016, 0.0080, -0.0006, ..., -0.0053, 0.0027, 0.0032], + dtype=torch.bfloat16), 'net.control_blocks_3.0.after_proj.weight': tensor([[-0.0035, 0.0068, 0.0042, ..., -0.0038, 0.0090, 0.0054], + [-0.0079, -0.0034, 0.0029, ..., -0.0081, 0.0088, -0.0048], + [ 0.0004, -0.0017, -0.0006, ..., -0.0027, 0.0028, -0.0023], + ..., + [ 0.0025, -0.0022, 0.0061, ..., 0.0028, 0.0021, -0.0047], + [ 0.0118, -0.0138, -0.0078, ..., -0.0090, -0.0025, 0.0232], + [ 0.0028, 0.0028, 0.0025, ..., 0.0020, -0.0011, 0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.after_proj.bias': tensor([ 0.0344, -0.0278, 0.0008, ..., -0.0009, -0.0157, 0.0364], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_proj.weight': tensor([[-0.0245, -0.0016, -0.0092, ..., -0.0034, 0.0038, -0.0042], + [-0.0108, -0.0238, -0.0061, ..., -0.0143, -0.0064, -0.0309], + [ 0.0024, 0.0037, -0.0109, ..., 0.0135, 0.0092, -0.0177], + ..., + [ 0.0058, 0.0115, 0.0014, ..., 0.0066, -0.0047, -0.0063], + [ 0.0062, 0.0038, 0.0052, ..., 0.0110, 0.0022, 0.0024], + [-0.0030, -0.0042, 0.0009, ..., -0.0003, -0.0014, 0.0046]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_norm.weight': tensor([ 1.6211e-01, 5.1953e-01, 9.4727e-02, 6.3672e-01, 5.7031e-01, + 4.7265e-08, -2.4872e-03, -2.4011e-10, 2.0339e-13, -1.2306e-28, + 1.0193e-02, 3.3881e-19, -1.1181e-19, -2.1546e-33, 3.7748e-15, + -2.8407e-33, 1.8066e-01, -9.9182e-04, 4.8065e-04, 7.3547e-03, + 1.1921e-05, 5.1117e-04, 1.9727e-01, 7.4219e-01, 7.6172e-01, + 3.8867e-01, -5.8350e-02, 4.3359e-01, -1.3428e-02, 1.6602e-02, + -2.4128e-04, 1.0991e-14, -1.3123e-02, 6.1989e-06, 5.6458e-04, + -3.1433e-03, 3.5390e-08, 2.6090e-15, 4.0531e-05, 4.1567e-13, + 8.4128e-12, -3.3329e-29, 1.7975e-07, 8.5156e-01, 2.4707e-01, + 2.4121e-01, 2.6562e-01, 7.4609e-01, 1.8555e-01, -4.4189e-02, + 1.1902e-02, 1.8359e-01, 1.4305e-04, -1.0315e-02, -1.3924e-04, + 8.0945e-11, 1.6846e-02, -6.3109e-29, 2.7008e-03, -9.2699e-18, + 9.9487e-03, 1.9560e-34, -1.4572e-15, -1.5247e-26, 3.1445e-01, + 1.9922e-01, 4.9414e-01, 1.5527e-01, 9.0332e-02, 2.5495e-08, + 1.1719e-02, 8.4401e-10, 3.4106e-13, -9.5526e-31, -6.5308e-03, + 1.3485e-18, 1.7047e-20, -3.7074e-33, -2.8644e-14, -3.8519e-34, + -5.4932e-03, -2.9206e-05, -1.2131e-03, 2.9755e-03, 6.2943e-05, + 4.2725e-02, 8.1641e-01, 2.4316e-01, 2.2852e-01, 7.5000e-01, + 2.5586e-01, 2.5391e-01, -1.4893e-02, 4.1809e-03, -1.3306e-02, + 2.4203e-14, 1.0193e-02, 2.1338e-05, 3.3140e-05, 3.3417e-03, + 1.4043e-09, -5.7554e-13, 2.3651e-03, -8.5348e-16, 2.1316e-14, + 2.6770e-32, -1.9222e-06, 2.1582e-01, 7.8125e-01, 7.2656e-01, + 7.3438e-01, 2.1973e-01, 3.7305e-01, -5.6076e-04, 2.4414e-01, + 1.1353e-02, 2.4986e-04, -8.4839e-03, -1.3428e-03, 7.3123e-10, + 2.1606e-02, -5.9165e-29, 3.3875e-03, -1.1350e-19, -6.1035e-04, + 1.2488e-34, -9.4369e-15, -1.3545e-23], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.self_attn.k_proj.weight': tensor([[ 0.0018, -0.0083, 0.0069, ..., -0.0095, -0.0078, -0.0113], + [ 0.0317, -0.0110, 0.0075, ..., -0.0096, -0.0066, -0.0208], + [ 0.0042, 0.0084, 0.0037, ..., 0.0049, 0.0028, 0.0038], + ..., + [ 0.0018, -0.0026, -0.0050, ..., -0.0085, -0.0002, 0.0064], + [-0.0094, 0.0029, 0.0021, ..., 0.0157, 0.0093, -0.0103], + [ 0.0045, -0.0045, -0.0019, ..., -0.0126, -0.0001, 0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.k_norm.weight': tensor([ 2.1973e-01, 5.3906e-01, -8.6914e-02, 6.4844e-01, 5.7422e-01, + -3.9581e-08, -8.8501e-03, 4.4929e-10, -1.4477e-13, -5.2854e-29, + -1.1414e-02, -5.6243e-19, -1.0927e-19, -5.6273e-34, 3.0809e-15, + -9.8704e-34, 1.7969e-01, -9.8419e-04, -4.8065e-04, 4.8584e-02, + 1.1861e-05, 7.0095e-05, 2.0508e-01, 7.5000e-01, 7.6562e-01, + 1.4746e-01, 1.9629e-01, 4.9414e-01, -1.1292e-02, 1.6479e-02, + 8.1787e-03, 1.7097e-14, 1.5259e-02, 3.3230e-06, 5.8365e-04, + 3.7842e-02, 3.3993e-08, 5.4401e-15, 4.0293e-05, 4.2100e-13, + -8.4697e-12, -3.3921e-29, -6.0350e-07, 8.4766e-01, 2.3438e-01, + 2.1210e-03, 6.4941e-02, 7.4609e-01, 9.1797e-02, 3.6774e-03, + -2.4609e-01, 7.7637e-02, 1.4019e-04, -9.6436e-03, -1.0300e-04, + 1.2915e-10, 1.7578e-02, -6.1531e-29, -2.4567e-03, -7.9689e-18, + 9.7656e-03, 7.8993e-35, -4.8017e-15, 5.9373e-26, 3.0469e-01, + 2.3535e-01, 4.9805e-01, 1.2988e-01, -9.4238e-02, -3.0734e-08, + -7.0496e-03, 3.2742e-10, 2.4691e-13, 1.4298e-29, 1.6403e-03, + 1.3891e-18, 1.8423e-20, -6.9213e-34, -2.8200e-14, -8.5764e-35, + 4.5776e-03, -4.0293e-05, -1.2054e-03, 2.8076e-03, 6.2466e-05, + 4.2969e-02, 8.1250e-01, 1.2988e-01, 1.7969e-01, 7.6953e-01, + 2.2656e-01, 2.3047e-01, 1.6235e-02, -3.2501e-03, 9.4604e-03, + -2.2760e-14, 5.4626e-03, -2.2054e-05, -4.7386e-06, 3.1494e-02, + 1.3461e-09, -5.2935e-13, -2.4109e-03, -2.1927e-15, 9.3814e-15, + 5.4696e-32, -1.9222e-06, 2.5781e-01, 7.8516e-01, 7.3047e-01, + 7.4219e-01, -5.1270e-02, 4.0625e-01, -4.3701e-02, 5.5542e-03, + -1.5527e-01, 2.5558e-04, 9.0332e-03, -1.3428e-03, -7.2396e-10, + -1.6846e-02, -6.0348e-29, -3.0823e-03, -1.8465e-19, -5.7220e-04, + 9.7049e-35, -8.5487e-15, -1.1529e-23], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.self_attn.v_proj.weight': tensor([[ 5.7068e-03, 4.6387e-03, 6.9275e-03, ..., -1.5198e-02, + 3.6621e-03, 9.0599e-05], + [ 6.2561e-03, -3.5553e-03, 1.4343e-02, ..., -2.3193e-03, + -2.3956e-03, 3.4523e-04], + [ 4.8828e-03, 4.3030e-03, 3.4790e-03, ..., -6.3782e-03, + 5.6763e-03, -7.4463e-03], + ..., + [ 8.7891e-03, -1.0834e-03, 1.0437e-02, ..., 8.3618e-03, + 3.5248e-03, -1.6724e-02], + [-6.5308e-03, 4.5166e-03, 6.0425e-03, ..., 3.4180e-02, + -2.7924e-03, 5.6458e-03], + [-3.3875e-03, 1.0757e-03, -2.8076e-03, ..., 1.3245e-02, + -4.1199e-03, 4.6692e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.output_proj.weight': tensor([[-5.7373e-03, 7.4768e-03, 7.9155e-05, ..., 1.1292e-02, + -1.1719e-02, 7.5073e-03], + [ 1.1047e-02, -3.5553e-03, 1.7578e-02, ..., 4.7913e-03, + -1.2207e-02, 9.5215e-03], + [ 1.9531e-03, 2.9564e-04, 2.3499e-03, ..., 7.7820e-04, + 1.1841e-02, 4.6692e-03], + ..., + [ 1.1047e-02, 4.1504e-03, 1.1749e-03, ..., -4.1809e-03, + -7.9956e-03, -7.2937e-03], + [-7.1106e-03, 8.9722e-03, 4.9438e-03, ..., 4.0588e-03, + -3.7079e-03, 3.6926e-03], + [-1.1475e-02, 9.1553e-03, 1.1597e-02, ..., -2.4719e-03, + 8.1177e-03, -6.1340e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_proj.weight': tensor([[-0.0058, 0.0116, 0.0141, ..., -0.0040, 0.0054, 0.0082], + [ 0.0051, -0.0135, -0.0170, ..., 0.0056, -0.0195, -0.0311], + [ 0.0070, 0.0134, -0.0043, ..., -0.0032, 0.0034, -0.0074], + ..., + [ 0.0020, 0.0212, -0.0030, ..., 0.0063, 0.0024, -0.0049], + [-0.0104, 0.0107, 0.0090, ..., -0.0048, 0.0112, 0.0060], + [ 0.0151, -0.0081, -0.0079, ..., -0.0070, -0.0042, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_norm.weight': tensor([ 1.9775e-02, 4.4727e-01, 2.8320e-02, 4.8578e-06, 3.5048e-05, + 3.2507e-13, 2.3651e-03, 5.1172e-01, 2.6855e-03, 3.7695e-01, + 3.6133e-02, 1.2512e-02, 6.3171e-03, 1.2695e-02, 2.7466e-02, + 1.5381e-02, 1.2207e-02, 7.8201e-05, 2.8163e-06, 3.7109e-02, + 2.5391e-01, 2.5879e-02, 2.4915e-05, 2.4414e-02, 9.4604e-03, + 5.1737e-05, 4.1962e-05, 4.5166e-02, 1.3542e-04, 3.8281e-01, + -4.8399e-05, -1.2684e-04, 5.1880e-03, 8.8215e-05, 2.7008e-03, + 3.2806e-03, 1.9651e-07, 4.1211e-01, 1.0315e-02, 5.4932e-03, + 6.9336e-02, 1.2268e-02, 1.6846e-02, 9.5844e-05, 1.1826e-03, + 1.4526e-02, 3.8910e-03, 2.2217e-02, 2.2461e-02, 1.9775e-02, + 1.1414e-02, 3.8281e-01, 1.2360e-03, 1.9653e-02, 2.6489e-02, + 3.3875e-03, 1.0205e-01, 1.0132e-02, 2.3926e-02, 1.2573e-02, + 1.6975e-04, 5.3101e-03, 1.2878e-02, -1.7738e-04, 5.6152e-03, + 5.5542e-03, -1.0133e-05, 3.3188e-04, 1.6235e-02, 7.9590e-02, + 6.0425e-03, 8.1177e-03, 6.0547e-02, 4.1260e-02, 6.9275e-03, + 5.2185e-03, 7.6771e-05, 2.5879e-02, 3.2227e-02, 2.6489e-02, + 3.9673e-03, 2.0508e-02, 3.1738e-02, -4.5868e-08, 1.9653e-02, + 1.3733e-02, 2.7895e-05, 4.5410e-02, 1.1475e-02, 3.4943e-03, + 1.3916e-02, 7.1335e-04, 5.0049e-03, 2.3193e-02, 1.2878e-02, + -1.3256e-04, 1.3245e-02, 4.0283e-02, 2.0264e-02, 4.8096e-02, + 3.9551e-02, 3.1128e-02, 1.7944e-02, 3.3140e-05, -4.7493e-04, + 3.2616e-04, 1.0864e-02, 2.0264e-02, 1.8799e-02, 1.9287e-02, + 3.7384e-03, 1.3977e-02, 3.6865e-02, 4.9414e-01, -1.6391e-06, + 2.5269e-02, 6.9580e-03, 1.5991e-02, 1.4572e-03, 1.1182e-01, + 5.0068e-06, 2.2411e-05, 3.4571e-05, 1.1475e-02, 1.5015e-02, + 1.5625e-02, 2.3804e-02, 3.1494e-02], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_proj.weight': tensor([[-0.0017, -0.0033, 0.0020, ..., -0.0039, 0.0030, 0.0041], + [ 0.0123, -0.0035, 0.0190, ..., -0.0018, 0.0317, 0.0050], + [ 0.0035, 0.0036, 0.0027, ..., 0.0023, 0.0099, -0.0018], + ..., + [-0.0064, -0.0049, -0.0075, ..., -0.0028, 0.0013, 0.0076], + [-0.0078, -0.0053, -0.0181, ..., -0.0019, 0.0045, 0.0051], + [-0.0024, -0.0004, -0.0070, ..., 0.0022, 0.0089, -0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_norm.weight': tensor([ 2.0142e-02, 4.4727e-01, 2.8198e-02, 4.8876e-06, 3.5048e-05, + -3.1086e-13, 2.3651e-03, 5.1172e-01, 2.7618e-03, 3.7695e-01, + 3.7109e-02, 1.2512e-02, 6.2866e-03, 1.2817e-02, 2.7588e-02, + 1.5259e-02, 1.2451e-02, 7.7248e-05, 3.7402e-06, 3.7354e-02, + 2.5391e-01, 2.5146e-02, 2.4319e-05, 2.4658e-02, 9.5215e-03, + 5.0783e-05, 4.4346e-05, 4.4678e-02, 1.3733e-04, 3.8477e-01, + -4.8637e-05, -1.2684e-04, 5.0659e-03, 8.7738e-05, 2.7313e-03, + 3.3112e-03, 2.0582e-07, 4.1211e-01, 1.0315e-02, 5.4626e-03, + 6.8359e-02, 1.2207e-02, 1.6724e-02, 9.5367e-05, 1.1063e-03, + 1.4526e-02, 3.9368e-03, 2.1851e-02, 2.2461e-02, 2.0020e-02, + 1.1475e-02, 3.8281e-01, 1.1444e-03, 1.9287e-02, 2.6123e-02, + 3.3112e-03, 1.0205e-01, 1.0132e-02, 2.3438e-02, 1.2512e-02, + 1.6785e-04, 5.3406e-03, 1.3000e-02, -1.7738e-04, 5.6152e-03, + 5.4626e-03, -8.5235e-06, 3.1662e-04, 1.6357e-02, 7.9590e-02, + 6.0120e-03, 8.1177e-03, 6.0059e-02, 4.1260e-02, 6.9885e-03, + 5.3406e-03, 7.7248e-05, 2.5513e-02, 3.1738e-02, 2.6489e-02, + 3.9673e-03, 2.0386e-02, 3.1250e-02, -4.4005e-08, 1.9531e-02, + 1.3672e-02, 2.8133e-05, 4.5410e-02, 1.1902e-02, 3.6774e-03, + 1.4282e-02, 7.1716e-04, 5.0049e-03, 2.3071e-02, 1.2756e-02, + -1.3447e-04, 1.2695e-02, 3.9795e-02, 2.0264e-02, 4.8096e-02, + 4.0283e-02, 3.1128e-02, 1.8188e-02, 3.4094e-05, 4.0054e-04, + 3.0518e-04, 1.0925e-02, 1.9775e-02, 1.8921e-02, 1.9165e-02, + 3.7689e-03, 1.3672e-02, 3.6621e-02, 4.9414e-01, -8.8289e-07, + 2.5146e-02, 6.9885e-03, 1.5869e-02, 1.6556e-03, 1.1230e-01, + 2.8312e-06, 2.1338e-05, 3.6478e-05, 1.1719e-02, 1.5076e-02, + 1.5747e-02, 2.3804e-02, 3.1738e-02], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.v_proj.weight': tensor([[-0.0183, 0.0031, 0.0049, ..., 0.0004, -0.0172, -0.0260], + [-0.0047, 0.0093, 0.0052, ..., 0.0046, 0.0053, 0.0051], + [-0.0417, 0.0022, -0.0008, ..., -0.0071, -0.0254, -0.0123], + ..., + [-0.0391, 0.0014, -0.0173, ..., 0.0022, 0.0209, 0.0052], + [ 0.0125, 0.0042, -0.0064, ..., -0.0048, 0.0092, 0.0037], + [ 0.0103, 0.0024, -0.0087, ..., -0.0036, -0.0029, 0.0072]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.output_proj.weight': tensor([[-6.5308e-03, -1.1597e-02, 2.0142e-02, ..., 1.0864e-02, + 1.2268e-02, 5.9509e-03], + [-5.8899e-03, -6.9885e-03, -4.7607e-03, ..., -2.9755e-03, + 5.1270e-03, 1.6724e-02], + [-2.5177e-03, 4.0588e-03, -6.8054e-03, ..., -2.5787e-03, + -2.1210e-03, 3.6926e-03], + ..., + [-3.4790e-03, 1.4954e-03, 6.2561e-03, ..., -6.8054e-03, + 6.3477e-03, 3.5553e-03], + [ 4.3640e-03, 4.1504e-03, 3.9368e-03, ..., -3.3203e-02, + -3.5706e-03, 3.3447e-02], + [ 1.1475e-02, 7.6294e-05, -1.5503e-02, ..., -1.7944e-02, + 3.6011e-03, -4.4556e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_img.weight': tensor([[ 0.0082, -0.0017, 0.0430, ..., -0.0327, -0.0052, 0.0155], + [-0.0278, 0.0280, 0.0249, ..., -0.0072, -0.0378, 0.0271], + [ 0.0134, -0.0121, 0.0312, ..., -0.0077, 0.0253, -0.0113], + ..., + [ 0.0245, -0.0150, 0.0041, ..., -0.0452, 0.0216, -0.0209], + [-0.0081, 0.0069, -0.0471, ..., -0.0106, 0.0157, -0.0082], + [ 0.0062, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0059, -0.0122, -0.0117], + [ 0.0137, 0.0060, -0.0188, ..., 0.0157, 0.0082, -0.0052], + [-0.0306, 0.0012, -0.0272, ..., 0.0100, 0.0337, -0.0050], + ..., + [-0.0040, -0.0046, 0.0044, ..., 0.0209, -0.0107, 0.0071], + [ 0.0193, 0.0210, 0.0123, ..., -0.0146, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0109, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img.weight': tensor([[ 0.0045, -0.0057, 0.0173, ..., 0.0081, 0.0356, 0.0217], + [-0.0498, 0.0062, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0104, 0.0033, -0.0139, ..., -0.0042, 0.0269, -0.0049], + ..., + [ 0.0010, 0.0126, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0073, -0.0162, ..., -0.0057, -0.0073, -0.0120], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0139, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.mlp.layer1.weight': tensor([[-0.0029, 0.0050, -0.0092, ..., 0.0106, 0.0013, -0.0006], + [-0.0026, 0.0053, 0.0047, ..., 0.0059, 0.0116, -0.0123], + [-0.0070, 0.0026, 0.0059, ..., -0.0065, 0.0070, 0.0033], + ..., + [ 0.0194, -0.0036, 0.0075, ..., 0.0078, -0.0006, 0.0038], + [-0.0242, -0.0052, -0.0021, ..., 0.0217, -0.0055, -0.0046], + [ 0.0086, -0.0250, 0.0101, ..., -0.0140, -0.0094, -0.0160]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.mlp.layer2.weight': tensor([[-0.0098, 0.0081, 0.0093, ..., 0.0049, -0.0040, 0.0064], + [-0.0070, 0.0113, 0.0073, ..., -0.0020, 0.0010, -0.0175], + [ 0.0007, 0.0038, -0.0064, ..., 0.0023, 0.0041, -0.0114], + ..., + [-0.0067, -0.0072, 0.0131, ..., -0.0116, -0.0244, 0.0087], + [ 0.0146, 0.0072, -0.0046, ..., -0.0078, 0.0026, -0.0067], + [-0.0046, 0.0062, 0.0068, ..., 0.0039, -0.0120, -0.0143]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0405, -0.0396, -0.0391, ..., 0.0050, 0.0053, 0.0052], + [ 0.0327, 0.0315, 0.0306, ..., -0.0037, -0.0036, -0.0037], + [-0.0417, -0.0398, -0.0396, ..., 0.0050, 0.0051, 0.0051], + ..., + [-0.0447, -0.0435, -0.0422, ..., 0.0034, 0.0036, 0.0037], + [-0.0386, -0.0381, -0.0371, ..., 0.0037, 0.0039, 0.0039], + [-0.0228, -0.0220, -0.0210, ..., 0.0012, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0043, -0.0006, 0.0045, ..., 0.0017, 0.0025, 0.0002], + [-0.0104, 0.0042, -0.0062, ..., -0.0077, -0.0066, -0.0042], + [-0.0049, 0.0018, -0.0061, ..., -0.0084, -0.0084, -0.0020], + ..., + [-0.0220, 0.0037, -0.0267, ..., -0.0286, -0.0270, -0.0038], + [-0.0124, -0.0055, -0.0131, ..., -0.0153, -0.0137, 0.0064], + [ 0.0064, 0.0004, 0.0054, ..., 0.0051, 0.0044, -0.0004]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9570e-02, 5.7861e-02, 5.7129e-02, ..., -1.2159e-04, + -1.6308e-04, -1.1539e-04], + [-4.0283e-03, -3.9978e-03, -3.9978e-03, ..., -4.1008e-05, + -1.1444e-04, -2.2888e-04], + [-1.8555e-02, -1.7944e-02, -1.7090e-02, ..., -6.9809e-04, + -7.0572e-04, -7.1716e-04], + ..., + [-2.9053e-02, -2.8931e-02, -2.8076e-02, ..., -4.1771e-04, + -3.7384e-04, -3.3569e-04], + [-3.3203e-02, -3.2227e-02, -3.0884e-02, ..., -1.8311e-04, + -1.3638e-04, -1.9073e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., -2.0504e-04, + -8.8692e-05, -2.3460e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0089, 0.0040, -0.0204, ..., 0.0052, 0.0135, -0.0127], + [ 0.0073, 0.0063, -0.0064, ..., 0.0061, -0.0039, 0.0053], + [ 0.0121, 0.0049, -0.0063, ..., 0.0080, -0.0047, 0.0102], + ..., + [ 0.0491, 0.0076, -0.0053, ..., 0.0062, -0.0149, 0.0476], + [ 0.0205, -0.0086, 0.0053, ..., -0.0145, -0.0229, 0.0243], + [-0.0029, 0.0096, -0.0164, ..., 0.0114, 0.0080, -0.0063]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_mlp.1.weight': tensor([[-0.0078, -0.0078, -0.0078, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0369, 0.0366, ..., -0.0038, -0.0038, -0.0037], + [ 0.0408, 0.0400, 0.0396, ..., -0.0038, -0.0038, -0.0039], + ..., + [ 0.0262, 0.0253, 0.0254, ..., -0.0021, -0.0021, -0.0020], + [ 0.0057, 0.0063, 0.0064, ..., 0.0014, 0.0014, 0.0013], + [ 0.0166, 0.0170, 0.0167, ..., -0.0025, -0.0025, -0.0026]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_mlp.2.weight': tensor([[-3.2654e-03, 6.0120e-03, 4.9133e-03, ..., 4.3640e-03, + 3.2959e-03, 3.2959e-03], + [ 7.0190e-03, 4.7607e-03, 2.6093e-03, ..., -1.2302e-04, + -7.4158e-03, 3.6240e-05], + [ 3.5553e-03, -4.0894e-03, -5.2490e-03, ..., -3.2806e-03, + -3.6621e-03, -3.2501e-03], + ..., + [ 4.0245e-04, 4.3335e-03, -4.7607e-03, ..., -5.8289e-03, + 3.7994e-03, 3.7537e-03], + [ 7.9956e-03, 1.6968e-02, 1.3916e-02, ..., -4.1504e-03, + -7.9956e-03, -3.8147e-03], + [-4.6387e-03, -4.7913e-03, 2.6855e-03, ..., 4.2114e-03, + 4.6387e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.after_proj.weight': tensor([[ 8.1635e-04, 1.4099e-02, 9.2163e-03, ..., -7.5073e-03, + 9.3994e-03, -5.6458e-03], + [-5.2490e-03, -1.0864e-02, -5.1498e-05, ..., 1.3306e-02, + -3.2959e-03, -6.2866e-03], + [ 3.9673e-03, -7.7820e-03, -8.8120e-04, ..., -5.7068e-03, + -1.7944e-02, 6.0120e-03], + ..., + [-5.3101e-03, 6.7749e-03, 2.5757e-02, ..., -2.0020e-02, + -7.4768e-03, -7.4768e-03], + [-1.1902e-02, -5.5237e-03, 1.6632e-03, ..., 1.1230e-02, + 1.0254e-02, -6.3782e-03], + [-4.4556e-03, 7.5378e-03, -4.5776e-03, ..., -6.2561e-03, + -5.9814e-03, 3.9673e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.after_proj.bias': tensor([ 0.0052, 0.0009, 0.0016, ..., 0.0064, 0.0066, -0.0004], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_proj.weight': tensor([[ 0.0437, -0.0240, -0.0189, ..., 0.0144, -0.0232, 0.0200], + [-0.0255, -0.0161, -0.0065, ..., -0.0056, 0.0101, 0.0036], + [ 0.0311, 0.0232, 0.0126, ..., -0.0045, -0.0199, -0.0047], + ..., + [-0.0085, -0.0048, 0.0018, ..., 0.0041, 0.0041, -0.0118], + [ 0.0167, 0.0259, 0.0025, ..., 0.0038, -0.0092, 0.0112], + [-0.0107, -0.0045, 0.0186, ..., -0.0060, 0.0145, -0.0156]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_norm.weight': tensor([ 1.2344e+00, 3.1836e-01, 4.5117e-01, 2.8711e-01, 1.1172e+00, + 7.7515e-03, 1.1841e-02, -4.2455e-13, 2.0874e-02, 4.4861e-03, + -5.4250e-08, 1.9096e-13, 1.0000e+00, -7.9632e-05, -1.8860e-22, + -2.9419e-02, -6.1035e-03, 8.8501e-03, 1.6235e-02, 1.3916e-02, + 1.5820e-01, -1.4782e-04, 5.0000e-01, 1.3359e+00, 6.4062e-01, + 7.3828e-01, 3.2812e-01, 1.6602e-01, 3.5156e-02, -1.2146e-02, + -3.3875e-03, 5.7459e-05, 5.9509e-03, 1.6785e-04, -6.7749e-03, + 3.3936e-02, 1.1816e-01, -9.1553e-05, 2.4780e-02, 1.1504e-05, + 1.4210e-04, -1.5974e-05, 4.5076e-07, 1.0469e+00, 1.2793e-01, + 1.1875e+00, 6.4062e-01, 1.4355e-01, -4.2969e-02, 1.1108e-02, + 1.0620e-02, 1.6846e-02, 1.3867e-01, -8.9645e-05, 1.6499e-04, + 1.0048e-14, 2.2602e-04, -9.8944e-06, -5.5176e-02, -3.8025e-15, + 3.0640e-02, 7.2718e-06, 5.7983e-03, -3.7231e-03, 5.8594e-01, + 1.1562e+00, 1.1328e+00, 1.1406e+00, 3.1250e-01, -1.2634e-02, + 8.2397e-03, 5.3957e-14, 2.1118e-02, -1.2085e-02, -7.3924e-09, + -5.5067e-14, 4.1504e-02, 2.7008e-03, -3.1764e-20, 6.7188e-01, + 9.1406e-01, -9.2773e-03, -5.6744e-05, 2.6611e-02, 1.1621e-01, + 7.7515e-03, 1.2812e+00, 5.3125e-01, 1.1875e+00, 1.2500e+00, + 2.4121e-01, 4.1992e-01, 4.4922e-02, -1.5869e-02, 4.3335e-03, + 5.0068e-05, -2.5269e-02, -3.8147e-03, 1.7578e-02, 3.5645e-02, + 1.2598e-01, -9.8705e-05, 9.7266e-01, -6.0536e-09, 1.2970e-04, + -1.3471e-05, -1.3279e-10, 6.9531e-01, 1.3750e+00, 1.1953e+00, + 1.2969e+00, 3.3789e-01, 2.5879e-02, 2.0142e-02, 2.1973e-02, + 2.7344e-02, 1.3574e-01, -1.0347e-04, 1.6594e-04, 1.6542e-14, + 1.5068e-04, -1.1206e-05, 1.0469e+00, -4.8572e-15, 3.5889e-02, + 1.0058e-06, 5.9375e-01, 1.0547e+00], dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.self_attn.k_proj.weight': tensor([[-0.0256, 0.0096, 0.0081, ..., 0.0226, -0.0024, 0.0109], + [ 0.0042, 0.0034, 0.0022, ..., -0.0066, 0.0105, -0.0008], + [-0.0291, 0.0134, -0.0076, ..., -0.0049, 0.0238, 0.0005], + ..., + [-0.0023, -0.0036, 0.0055, ..., -0.0068, -0.0052, 0.0079], + [ 0.0024, 0.0055, 0.0126, ..., 0.0192, -0.0034, -0.0075], + [ 0.0066, 0.0018, 0.0126, ..., -0.0008, 0.0061, 0.0229]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.k_norm.weight': tensor([ 1.2812e+00, 7.1875e-01, 4.3555e-01, 2.9297e-01, 1.1172e+00, + 1.2939e-02, 1.9653e-02, 2.4158e-13, 2.1484e-02, 7.6904e-03, + -6.1002e-08, 1.9895e-13, 1.0000e+00, -7.9632e-05, -7.0310e-23, + -7.6904e-03, 2.6855e-03, 8.0078e-02, 1.8066e-02, 3.1586e-03, + 5.7678e-03, -1.0490e-04, 5.3125e-01, 1.2188e+00, 2.1191e-01, + 5.8594e-01, 4.0430e-01, 1.5625e-01, 1.0193e-02, -1.8692e-03, + 4.6997e-03, 4.8161e-05, 4.1199e-03, 4.0283e-02, 4.8218e-03, + -4.3640e-03, 4.3335e-03, -3.9816e-05, 2.9907e-02, 1.1981e-05, + 1.6880e-04, -1.2636e-05, -4.5449e-07, 1.1406e+00, 7.5781e-01, + 1.1406e+00, 4.4141e-01, 2.2070e-01, 1.6846e-02, 5.6641e-02, + -9.8267e-03, 2.8992e-03, 1.1414e-02, -2.9683e-05, 6.2943e-05, + -1.3600e-14, 2.4605e-04, -3.6210e-06, -2.2217e-02, -3.5388e-15, + -3.8452e-03, -4.6492e-06, 3.6621e-03, 1.3733e-02, 6.2891e-01, + 1.0859e+00, 1.1484e+00, 1.1172e+00, 2.8516e-01, 1.8311e-03, + -2.2125e-03, -2.4691e-13, 1.6357e-02, 1.8539e-03, -7.6252e-09, + -4.5075e-14, -1.3184e-02, 2.7008e-03, 3.1764e-20, 6.7188e-01, + 9.1406e-01, 6.5430e-02, 1.5926e-04, 3.8147e-03, 8.8501e-03, + 7.9956e-03, 1.2109e+00, 7.9688e-01, 1.3047e+00, 1.2969e+00, + 2.5146e-02, 4.2188e-01, 1.7578e-02, 2.1362e-02, -3.4027e-03, + 3.6001e-05, 2.7954e-02, 4.5410e-02, 2.3193e-02, -8.1177e-03, + -1.8311e-02, -6.0320e-05, 9.7266e-01, 2.7707e-08, 1.6117e-04, + -9.1195e-06, 7.1907e-12, 4.7070e-01, 1.2188e+00, 1.1484e+00, + 1.4141e+00, 3.3789e-01, -3.8330e-02, 1.0205e-01, 2.1729e-02, + 1.2024e-02, 1.1475e-02, -5.6267e-05, 1.5354e-04, -2.2871e-14, + 1.7929e-04, -6.0499e-06, 1.0469e+00, -5.3291e-15, -1.5625e-02, + 1.2964e-06, 5.9375e-01, 1.0547e+00], dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0095, -0.0096, ..., -0.0177, 0.0064, 0.0099], + [-0.0042, -0.0049, -0.0126, ..., -0.0092, 0.0050, 0.0063], + [-0.0137, -0.0091, 0.0016, ..., 0.0320, -0.0079, -0.0100], + ..., + [ 0.0110, 0.0151, -0.0223, ..., -0.0342, -0.0092, -0.0047], + [ 0.0007, 0.0009, 0.0025, ..., 0.0045, 0.0035, -0.0098], + [ 0.0040, -0.0181, -0.0162, ..., 0.0044, -0.0074, -0.0038]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.output_proj.weight': tensor([[ 0.0148, 0.0173, -0.0488, ..., 0.0023, -0.0142, 0.0139], + [-0.0038, -0.0066, -0.0146, ..., 0.0210, 0.0125, -0.0125], + [ 0.0003, 0.0044, 0.0043, ..., 0.0037, 0.0005, 0.0061], + ..., + [ 0.0070, 0.0051, -0.0271, ..., 0.0037, 0.0056, 0.0059], + [ 0.0009, 0.0048, 0.0065, ..., 0.0038, 0.0160, -0.0135], + [ 0.0040, 0.0108, 0.0039, ..., -0.0024, 0.0085, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_proj.weight': tensor([[-0.0053, 0.0062, -0.0029, ..., 0.0040, 0.0022, 0.0024], + [ 0.0075, -0.0033, 0.0025, ..., 0.0078, 0.0012, 0.0060], + [ 0.0046, 0.0026, -0.0014, ..., -0.0047, 0.0080, -0.0036], + ..., + [-0.0021, 0.0029, -0.0014, ..., -0.0053, 0.0069, -0.0017], + [-0.0088, -0.0069, 0.0020, ..., 0.0078, -0.0011, 0.0081], + [-0.0026, -0.0013, -0.0044, ..., -0.0045, 0.0035, -0.0061]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_norm.weight': tensor([ 2.8076e-03, 1.4062e-01, 2.1118e-02, 6.1719e-01, 2.1240e-02, + 1.4526e-02, 1.2695e-02, 5.3406e-03, 2.0386e-02, 1.2305e-01, + 1.7212e-02, 5.6641e-02, 2.6758e-01, 2.2217e-02, 1.7969e-01, + 2.2461e-02, 6.3965e-02, 2.4292e-02, 5.4016e-03, 2.3438e-02, + 1.0681e-02, 1.6895e-01, 2.6562e-01, 2.0410e-01, 2.4902e-02, + 1.5820e-01, 5.5420e-02, 1.1475e-02, 3.8867e-01, 1.2695e-02, + 1.5015e-02, 5.0781e-01, 4.6631e-02, 4.3701e-02, 9.1553e-03, + 2.0020e-02, 8.9355e-02, 5.0391e-01, 5.7422e-01, 1.1841e-02, + 1.1414e-02, 1.0059e-01, 5.2795e-03, 5.5469e-01, 2.9297e-02, + 4.3213e-02, 1.1169e-02, 3.9648e-01, 8.3618e-03, 2.6978e-02, + 6.7871e-02, 1.5747e-02, 2.0264e-02, 5.1953e-01, 2.7313e-03, + 2.8076e-02, 4.7119e-02, 6.7871e-02, 1.1719e-02, 2.0752e-02, + 3.1281e-03, 1.4587e-02, 5.2490e-03, 3.6328e-01, 2.3193e-02, + 1.7700e-03, 2.4902e-02, 6.4453e-02, 5.5664e-02, 9.1553e-03, + 9.4604e-03, 6.2109e-01, -3.6716e-05, 6.0547e-01, 2.0386e-02, + 4.1962e-04, 1.4587e-02, 3.1055e-01, 2.9297e-02, 4.5586e-04, + 8.1055e-02, 8.2397e-03, 5.2246e-02, 9.8877e-03, 1.6235e-02, + 8.0566e-03, 1.1719e-02, 5.1025e-02, 2.0996e-01, 1.1536e-02, + 1.0437e-02, 2.3560e-02, 4.2480e-02, 3.3447e-02, 3.0640e-02, + 1.7456e-02, 1.3550e-02, 2.3560e-02, 5.0537e-02, 2.7466e-02, + 5.0049e-02, 1.9897e-02, 6.2561e-03, 5.2246e-02, 5.9766e-01, + 6.0547e-02, 7.2021e-03, 4.6875e-01, 1.5015e-02, 2.8931e-02, + 1.2085e-02, 1.4587e-02, 7.8201e-05, 7.9102e-02, 4.5117e-01, + 2.0264e-02, 2.4048e-02, 5.6641e-01, 1.0254e-01, 3.1128e-02, + 1.5015e-02, 6.9531e-01, 1.4062e-01, 1.6235e-02, 1.1169e-02, + 7.8125e-03, 2.2461e-02, 1.2634e-02], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_proj.weight': tensor([[ 0.0013, -0.0034, 0.0126, ..., -0.0002, -0.0052, 0.0074], + [-0.0046, 0.0003, -0.0056, ..., 0.0067, -0.0096, -0.0072], + [-0.0070, -0.0016, -0.0056, ..., 0.0168, -0.0151, -0.0118], + ..., + [-0.0018, 0.0046, -0.0001, ..., -0.0026, -0.0044, -0.0068], + [ 0.0043, 0.0029, -0.0074, ..., -0.0056, 0.0054, 0.0016], + [-0.0072, 0.0007, 0.0049, ..., 0.0015, 0.0039, 0.0058]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_norm.weight': tensor([ 2.8381e-03, 1.4062e-01, 2.0996e-02, 6.1719e-01, 2.0996e-02, + 1.4526e-02, 1.2817e-02, 5.2185e-03, 2.0630e-02, 1.2305e-01, + 1.7212e-02, 5.5176e-02, 2.6758e-01, 2.2095e-02, 1.7969e-01, + 2.3193e-02, 6.3477e-02, 2.4414e-02, 5.3406e-03, 2.3315e-02, + 1.0986e-02, 1.6992e-01, 2.6758e-01, 2.0410e-01, 2.4780e-02, + 1.5820e-01, 5.4443e-02, 9.3994e-03, 3.9062e-01, 1.2817e-02, + 1.5198e-02, 5.0781e-01, 4.6631e-02, 4.3701e-02, 9.5215e-03, + 1.9653e-02, 8.8867e-02, 5.0391e-01, 5.7422e-01, 1.1780e-02, + 1.1414e-02, 1.0010e-01, 5.2490e-03, 5.5469e-01, 2.9541e-02, + 4.3213e-02, 1.1475e-02, 3.9648e-01, 8.4839e-03, 2.6855e-02, + 6.7871e-02, 1.6357e-02, 2.0386e-02, 5.1953e-01, 3.1128e-03, + 2.7954e-02, 4.6387e-02, 6.7871e-02, 1.1780e-02, 1.9287e-02, + 3.0060e-03, 1.4526e-02, 5.6152e-03, 3.6328e-01, 2.2949e-02, + 1.9302e-03, 2.4414e-02, 6.4453e-02, 5.5664e-02, 9.0942e-03, + 9.8877e-03, 6.2109e-01, -5.1498e-05, 6.0547e-01, 2.0508e-02, + -1.7548e-04, 1.4648e-02, 3.0859e-01, 3.0029e-02, 6.9046e-04, + 8.1543e-02, 8.1787e-03, 5.1514e-02, 9.7656e-03, 1.5747e-02, + 8.1787e-03, 1.2329e-02, 5.1758e-02, 2.0898e-01, 1.1414e-02, + 1.0498e-02, 2.2583e-02, 4.2725e-02, 3.3203e-02, 3.1494e-02, + 1.7456e-02, 1.3611e-02, 2.3560e-02, 5.0537e-02, 2.7344e-02, + 4.9561e-02, 2.0020e-02, 6.8054e-03, 5.2734e-02, 5.9766e-01, + 6.0547e-02, 7.1411e-03, 4.6875e-01, 1.4587e-02, 2.8809e-02, + 1.2756e-02, 1.4587e-02, 8.0109e-05, 7.9102e-02, 4.5117e-01, + 1.9897e-02, 2.4048e-02, 5.6641e-01, 1.0352e-01, 3.1250e-02, + 1.4771e-02, 6.9922e-01, 1.3965e-01, 1.5259e-02, 1.0803e-02, + 7.6599e-03, 2.2339e-02, 1.2390e-02], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.v_proj.weight': tensor([[ 0.0130, -0.0059, -0.0171, ..., 0.0042, -0.0138, -0.0076], + [-0.0179, -0.0020, -0.0033, ..., -0.0018, -0.0227, 0.0190], + [-0.0127, -0.0049, 0.0033, ..., -0.0045, 0.0310, 0.0281], + ..., + [-0.0030, -0.0030, -0.0121, ..., 0.0048, 0.0171, 0.0043], + [ 0.0278, 0.0032, 0.0175, ..., 0.0035, 0.0116, 0.0164], + [ 0.0128, -0.0001, 0.0049, ..., 0.0053, -0.0192, -0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.output_proj.weight': tensor([[ 0.0125, -0.0001, -0.0154, ..., -0.0190, 0.0078, -0.0012], + [ 0.0089, 0.0038, 0.0125, ..., 0.0066, -0.0090, 0.0024], + [ 0.0059, -0.0033, -0.0007, ..., 0.0007, -0.0078, -0.0057], + ..., + [-0.0114, 0.0038, 0.0079, ..., 0.0052, -0.0232, 0.0073], + [ 0.0097, -0.0165, 0.0049, ..., -0.0265, -0.0133, -0.0056], + [ 0.0042, -0.0167, 0.0177, ..., 0.0317, -0.0025, 0.0086]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0065, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [-0.0032, 0.0150, -0.0159, ..., -0.0039, -0.0054, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0071, -0.0067], + [-0.0337, -0.0040, 0.0054, ..., -0.0041, 0.0172, 0.0231], + [-0.0255, -0.0040, -0.0039, ..., 0.0408, 0.0430, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0017], + [ 0.0198, -0.0208, 0.0128, ..., 0.0090, -0.0337, -0.0050], + [ 0.0302, 0.0101, -0.0029, ..., -0.0304, 0.0199, 0.0045], + ..., + [ 0.0171, -0.0276, -0.0073, ..., 0.0231, 0.0182, -0.0040], + [ 0.0221, 0.0388, 0.0056, ..., 0.0023, -0.0488, -0.0361], + [ 0.0058, 0.0118, -0.0055, ..., 0.0042, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img.weight': tensor([[ 2.7344e-02, -1.9908e-05, 1.1841e-02, ..., 1.4725e-03, + 3.4668e-02, 6.1523e-02], + [-2.5757e-02, 2.6703e-03, -1.9653e-02, ..., 1.3916e-02, + -4.6082e-03, 1.0986e-02], + [-8.2397e-03, -1.6846e-02, -4.2725e-02, ..., -5.8594e-02, + 1.5320e-02, 8.5449e-03], + ..., + [ 9.5825e-03, 1.3184e-02, 1.8433e-02, ..., 1.1047e-02, + 8.9111e-03, 5.1514e-02], + [-3.2806e-03, -2.4658e-02, -2.7832e-02, ..., -8.9722e-03, + -3.8818e-02, 8.3618e-03], + [ 1.9043e-02, -2.6733e-02, -2.1973e-02, ..., -4.5654e-02, + -1.5991e-02, 1.6113e-02]], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.mlp.layer1.weight': tensor([[-0.0068, 0.0051, 0.0058, ..., 0.0002, -0.0018, -0.0094], + [-0.0115, 0.0060, 0.0082, ..., 0.0094, -0.0030, 0.0022], + [ 0.0054, -0.0074, 0.0063, ..., 0.0050, -0.0054, -0.0349], + ..., + [-0.0066, -0.0027, 0.0054, ..., 0.0137, 0.0046, -0.0133], + [ 0.0049, 0.0027, -0.0034, ..., -0.0194, -0.0060, -0.0376], + [ 0.0045, 0.0277, 0.0002, ..., -0.0041, 0.0078, -0.0072]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.mlp.layer2.weight': tensor([[ 0.0056, 0.0057, -0.0063, ..., -0.0060, 0.0172, -0.0059], + [-0.0040, -0.0027, -0.0060, ..., 0.0073, -0.0140, -0.0192], + [ 0.0028, 0.0055, 0.0098, ..., 0.0089, 0.0079, -0.0041], + ..., + [-0.0043, 0.0032, -0.0209, ..., -0.0107, 0.0076, -0.0089], + [-0.0117, 0.0106, 0.0123, ..., -0.0101, 0.0023, -0.0024], + [-0.0011, 0.0063, -0.0120, ..., 0.0049, -0.0094, -0.0374]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0347, 0.0344, 0.0330, ..., -0.0041, -0.0042, -0.0043], + [-0.0334, -0.0322, -0.0317, ..., 0.0043, 0.0047, 0.0046], + [-0.0349, -0.0344, -0.0332, ..., 0.0045, 0.0048, 0.0048], + ..., + [-0.0393, -0.0381, -0.0374, ..., 0.0022, 0.0023, 0.0024], + [-0.0400, -0.0393, -0.0383, ..., 0.0021, 0.0022, 0.0022], + [ 0.0078, 0.0078, 0.0073, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0079, -0.0075, -0.0080, ..., -0.0095, -0.0101, 0.0018], + [ 0.0047, -0.0041, -0.0047, ..., -0.0023, -0.0013, 0.0017], + [ 0.0089, -0.0084, -0.0087, ..., -0.0045, -0.0043, -0.0007], + ..., + [ 0.0383, -0.0422, -0.0413, ..., -0.0222, -0.0182, 0.0082], + [ 0.0093, -0.0092, -0.0087, ..., -0.0067, -0.0067, -0.0113], + [-0.0181, 0.0188, 0.0179, ..., 0.0146, 0.0138, 0.0118]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9326e-02, 5.8350e-02, 5.7373e-02, ..., -1.2589e-04, + -5.5313e-05, -2.0790e-04], + [ 6.0547e-02, 5.8838e-02, 5.7129e-02, ..., 3.7575e-04, + 4.8828e-04, 5.1880e-04], + [-4.9561e-02, -4.7852e-02, -4.6631e-02, ..., -1.2934e-05, + 4.2439e-05, -6.6280e-05], + ..., + [ 3.5156e-02, 3.4180e-02, 3.3447e-02, ..., 1.0538e-04, + 3.7432e-05, 6.6757e-05], + [ 6.2988e-02, 6.2012e-02, 6.0791e-02, ..., 5.6076e-04, + 5.9891e-04, 5.9128e-04], + [-5.1758e-02, -5.1270e-02, -5.0049e-02, ..., -3.5286e-04, + -3.9864e-04, -3.5858e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0064, 0.0097, -0.0022, ..., -0.0041, -0.0103, 0.0048], + [-0.0140, 0.0099, 0.0109, ..., -0.0084, -0.0104, 0.0062], + [ 0.0036, 0.0033, -0.0029, ..., 0.0036, -0.0033, 0.0033], + ..., + [ 0.0182, 0.0115, -0.0118, ..., 0.0065, 0.0120, 0.0056], + [ 0.0114, 0.0291, -0.0128, ..., 0.0134, 0.0383, -0.0232], + [-0.0200, 0.0195, 0.0177, ..., -0.0133, -0.0099, 0.0231]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_mlp.1.weight': tensor([[-0.0757, -0.0742, -0.0718, ..., 0.0029, 0.0032, 0.0031], + [-0.0258, -0.0243, -0.0233, ..., 0.0006, 0.0007, 0.0006], + [ 0.0559, 0.0542, 0.0530, ..., -0.0017, -0.0017, -0.0017], + ..., + [ 0.0430, 0.0420, 0.0415, ..., -0.0019, -0.0018, -0.0016], + [ 0.0347, 0.0330, 0.0317, ..., -0.0009, -0.0008, -0.0009], + [-0.0520, -0.0505, -0.0496, ..., 0.0022, 0.0022, 0.0021]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_mlp.2.weight': tensor([[-0.0266, -0.0076, 0.0162, ..., 0.0093, 0.0091, -0.0166], + [ 0.0042, -0.0098, -0.0094, ..., -0.0092, 0.0098, 0.0092], + [-0.0038, -0.0087, -0.0049, ..., -0.0069, 0.0090, 0.0049], + ..., + [ 0.0608, -0.0090, -0.0262, ..., -0.0195, 0.0083, 0.0270], + [ 0.0187, -0.0172, -0.0186, ..., -0.0171, 0.0171, 0.0179], + [ 0.0054, -0.0179, -0.0103, ..., -0.0129, 0.0184, 0.0098]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.after_proj.weight': tensor([[-0.0054, 0.0061, -0.0078, ..., -0.0073, -0.0055, -0.0062], + [-0.0144, -0.0081, 0.0093, ..., 0.0078, -0.0068, 0.0075], + [ 0.0061, 0.0078, 0.0070, ..., -0.0076, 0.0082, -0.0051], + ..., + [ 0.0074, -0.0090, 0.0143, ..., 0.0154, 0.0192, 0.0157], + [ 0.0014, 0.0089, 0.0025, ..., -0.0200, -0.0118, 0.0017], + [-0.0080, -0.0099, -0.0138, ..., -0.0076, -0.0044, -0.0057]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.after_proj.bias': tensor([ 0.0055, -0.0029, 0.0007, ..., 0.0034, 0.0079, 0.0055], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_proj.weight': tensor([[-0.0192, 0.0198, 0.0232, ..., -0.0031, -0.0097, -0.0026], + [-0.0106, -0.0056, -0.0008, ..., 0.0075, -0.0066, -0.0070], + [-0.0182, -0.0057, 0.0047, ..., -0.0069, -0.0079, 0.0070], + ..., + [-0.0043, 0.0033, -0.0081, ..., -0.0080, 0.0104, 0.0029], + [ 0.0066, -0.0008, -0.0037, ..., -0.0289, -0.0199, -0.0289], + [ 0.0036, -0.0161, 0.0070, ..., 0.0107, -0.0027, -0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_norm.weight': tensor([ 9.3750e-01, 1.4609e+00, 1.3750e+00, 1.3594e+00, 1.3574e-01, + 5.8350e-02, 1.6479e-02, -2.1770e-08, -2.4435e-33, 3.0823e-03, + 5.4016e-03, 1.0840e-01, -6.6223e-03, -1.0315e-02, 1.9989e-03, + 1.4687e-04, 5.9326e-02, 9.1406e-01, 1.0156e+00, 3.9291e-04, + 1.4400e-04, 1.1406e+00, 1.6562e+00, 1.6406e+00, 1.0000e+00, + 8.8672e-01, 1.2969e+00, -1.0376e-02, 4.2439e-05, 6.1989e-05, + -9.3994e-03, 2.6131e-04, 2.0874e-02, 1.0078e+00, 9.3384e-03, + 1.0312e+00, 2.8076e-03, 8.0109e-05, 2.3438e-02, -8.1177e-03, + 5.2527e-07, 1.2207e-04, 1.9908e-05, 1.6797e+00, 1.2266e+00, + 1.2734e+00, 1.3438e+00, 1.4531e+00, 6.7871e-02, 1.7452e-04, + 2.9053e-02, 2.6172e-01, 1.0469e+00, 9.0820e-02, 5.1737e-05, + 1.0596e-01, 2.3804e-03, 1.1621e-01, 2.5630e-05, 6.9580e-03, + 1.9372e-06, -1.8845e-03, 3.7231e-03, 1.0938e+00, 1.5234e+00, + 6.7969e-01, 8.3594e-01, 1.1406e+00, 6.0303e-02, -3.1738e-02, + 3.4637e-03, -4.9127e-08, -3.9482e-33, 2.0752e-02, 1.4709e-02, + 2.5482e-03, 3.1281e-04, -2.6093e-03, -1.6724e-02, 1.3062e-02, + 5.8594e-02, -2.8076e-03, 1.1353e-02, 2.4986e-04, 1.5640e-04, + -6.5002e-03, 6.8359e-01, 6.7578e-01, 1.3281e+00, 1.3438e+00, + 7.0312e-01, 1.0681e-03, -3.3140e-05, 3.1471e-05, 4.6387e-03, + -4.5967e-04, 2.9297e-02, 1.6113e-01, 9.0332e-03, 6.4453e-02, + 1.9932e-04, 3.4809e-05, 1.2207e-01, -7.6771e-05, 5.1498e-05, + -2.6226e-05, 5.1260e-05, 4.9805e-01, 1.4453e+00, 1.3047e+00, + 8.4375e-01, 8.4375e-01, 7.3730e-02, -1.7166e-03, 2.7832e-02, + 1.0547e+00, 5.9570e-02, 9.8047e-01, 4.8161e-05, 2.8839e-03, + 2.6245e-02, 5.2490e-02, 1.2815e-05, 5.5695e-04, 9.3937e-05, + 2.3804e-03, 1.0156e+00, -1.0071e-02], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.self_attn.k_proj.weight': tensor([[-0.0176, -0.0067, -0.0075, ..., -0.0074, 0.0067, -0.0098], + [-0.0050, -0.0165, 0.0178, ..., -0.0085, -0.0135, 0.0146], + [-0.0064, 0.0046, 0.0004, ..., -0.0054, -0.0054, 0.0057], + ..., + [-0.0082, 0.0190, -0.0006, ..., 0.0033, 0.0077, 0.0007], + [ 0.0146, -0.0021, 0.0188, ..., -0.0072, 0.0193, -0.0062], + [-0.0028, -0.0042, 0.0037, ..., -0.0037, 0.0043, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.k_norm.weight': tensor([ 7.0312e-01, 1.4141e+00, 1.3672e+00, 1.3047e+00, 1.3672e-01, + 6.5918e-02, 1.1108e-02, -1.8976e-08, -1.8296e-33, -6.6376e-04, + 5.2185e-03, 1.0938e-01, -1.0681e-02, 3.4668e-02, -1.7929e-03, + 5.9843e-05, 3.9673e-03, 9.1406e-01, 1.0156e+00, 4.6921e-04, + 8.9645e-05, 1.1406e+00, 1.4844e+00, 1.4844e+00, 1.6235e-02, + 2.8516e-01, 1.3281e+00, 1.9302e-03, 4.5538e-05, -5.1498e-05, + -8.7280e-03, 5.8746e-04, -1.9287e-02, 1.0000e+00, 6.2561e-03, + 1.0234e+00, 2.6855e-03, 8.8692e-05, -2.5146e-02, -8.1787e-03, + -1.0505e-06, -1.2589e-04, 1.0550e-05, 1.3203e+00, 5.5859e-01, + 1.2031e+00, 1.3984e+00, 1.4297e+00, -4.9438e-03, -7.7820e-04, + 2.4719e-03, -2.2095e-02, 1.0469e+00, 6.2256e-02, -3.4809e-05, + 1.0596e-01, -1.2970e-03, 1.1621e-01, -7.5400e-06, 7.0801e-02, + -1.7583e-06, 2.1606e-02, -2.1484e-02, 1.0938e+00, 1.5625e+00, + 1.0312e+00, 7.8125e-01, 1.1484e+00, 4.9072e-02, -1.9775e-02, + -1.8799e-02, 5.2620e-08, 4.7426e-33, 2.1240e-02, 2.5391e-02, + 3.3936e-02, 7.2098e-04, 2.7344e-02, -3.0151e-02, -1.3123e-02, + -2.6855e-02, 2.8381e-03, 2.7618e-03, 3.1853e-04, 9.2506e-05, + 5.7678e-03, 1.1875e+00, 1.1484e+00, 1.4297e+00, 1.4141e+00, + 3.0859e-01, 1.1536e-02, -6.0320e-05, -4.3631e-05, 6.0425e-03, + -2.2507e-04, 1.2695e-02, 3.2806e-03, -8.5449e-03, -2.2095e-02, + 9.9182e-05, 5.6267e-05, 1.2158e-01, 3.9339e-05, 6.2943e-05, + 2.7657e-05, 4.7445e-05, 1.2344e+00, 1.6484e+00, 1.2734e+00, + 2.3730e-01, 7.7734e-01, -6.0120e-03, -1.5182e-03, -1.5320e-02, + 1.0547e+00, -2.9053e-02, 9.8438e-01, 4.4823e-05, 7.5989e-03, + 2.6489e-02, 5.2002e-02, 8.0466e-06, 4.1748e-02, 9.4414e-05, + 1.5015e-02, 1.0156e+00, -1.9302e-03], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.self_attn.v_proj.weight': tensor([[ 3.8147e-03, -6.7444e-03, 1.0315e-02, ..., 4.6631e-02, + -9.6893e-04, 2.5330e-03], + [ 1.1963e-02, 2.1973e-02, -8.9722e-03, ..., -2.1240e-02, + -8.8501e-03, -5.7678e-03], + [-2.0294e-03, 1.8677e-02, -4.8218e-03, ..., 1.5625e-02, + -4.4556e-03, 9.4604e-03], + ..., + [ 4.2200e-05, -4.1199e-03, -1.9531e-02, ..., 9.4604e-03, + 1.5381e-02, 3.2349e-03], + [ 2.8229e-03, 1.8433e-02, -5.8899e-03, ..., 3.9978e-03, + 4.2114e-03, -3.4943e-03], + [-9.0332e-03, -8.6060e-03, 5.1880e-03, ..., -6.7139e-03, + -8.2397e-03, -7.4463e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.output_proj.weight': tensor([[ 0.0072, 0.0190, 0.0056, ..., 0.0101, 0.0192, -0.0095], + [ 0.0275, 0.0043, -0.0049, ..., -0.0041, -0.0270, -0.0094], + [ 0.0039, 0.0334, -0.0115, ..., -0.0043, -0.0165, 0.0060], + ..., + [-0.0381, -0.0177, 0.0038, ..., 0.0108, -0.0082, -0.0183], + [ 0.0012, 0.0125, -0.0266, ..., -0.0176, 0.0238, -0.0033], + [-0.0432, 0.0046, 0.0233, ..., 0.0222, -0.0132, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_proj.weight': tensor([[-0.0003, -0.0081, -0.0045, ..., -0.0159, -0.0014, 0.0152], + [ 0.0070, -0.0017, -0.0037, ..., 0.0017, 0.0044, 0.0084], + [ 0.0084, 0.0048, -0.0092, ..., 0.0050, 0.0126, -0.0039], + ..., + [ 0.0060, 0.0171, 0.0045, ..., 0.0056, -0.0106, 0.0049], + [-0.0036, 0.0073, -0.0050, ..., -0.0056, -0.0114, 0.0037], + [ 0.0004, -0.0054, 0.0040, ..., -0.0049, 0.0120, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_norm.weight': tensor([0.1113, 0.0237, 0.4238, 0.0752, 0.1670, 0.0310, 0.1553, 0.1206, 0.5195, + 0.0447, 0.0654, 0.0161, 0.0518, 0.0674, 0.0640, 0.2021, 0.4570, 0.0344, + 0.7539, 0.0364, 0.0327, 0.1689, 0.0193, 0.0894, 0.0623, 0.4355, 0.5234, + 0.2754, 0.0767, 0.0349, 0.0139, 0.2305, 0.6719, 0.2539, 0.0466, 0.0271, + 0.0216, 0.0967, 0.0184, 0.0383, 0.0527, 0.1201, 0.0229, 0.0439, 0.2080, + 0.0498, 0.3555, 0.0260, 0.0554, 0.0211, 0.0258, 0.4023, 0.5312, 0.3789, + 0.1025, 0.5898, 0.5938, 0.0366, 0.1367, 0.0144, 0.0913, 0.4805, 0.0153, + 0.5273, 0.1699, 0.4160, 0.1016, 0.0498, 0.0732, 0.4277, 0.0281, 0.5508, + 0.0251, 0.0442, 0.1533, 0.0908, 0.0243, 0.1113, 0.6406, 0.0398, 0.0952, + 0.0879, 0.0292, 0.0461, 0.4219, 0.0674, 0.0708, 0.0854, 0.0172, 0.1533, + 0.0432, 0.1787, 0.0471, 0.0703, 0.0391, 0.1201, 0.0114, 0.3555, 0.5391, + 0.0503, 0.0549, 0.3242, 0.0339, 0.1309, 0.0398, 0.2285, 0.0156, 0.5742, + 0.5273, 0.0359, 0.0171, 0.7344, 0.0273, 0.0225, 0.4746, 0.2178, 0.3008, + 0.7227, 0.0369, 0.1001, 0.0801, 0.5625, 0.6055, 0.0199, 0.0188, 0.1914, + 0.0374, 0.0408], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_proj.weight': tensor([[-1.9531e-03, 1.8158e-03, 6.0730e-03, ..., -1.7090e-03, + 1.0925e-02, -1.0864e-02], + [ 5.8289e-03, -6.7711e-05, -1.7776e-03, ..., 1.7548e-03, + -3.8452e-03, 1.2390e-02], + [-1.0437e-02, 7.6904e-03, -1.3428e-02, ..., 1.0803e-02, + 5.3406e-03, 4.2114e-03], + ..., + [ 7.5378e-03, 3.1128e-03, -7.1411e-03, ..., -4.4250e-04, + -1.6724e-02, 5.8594e-03], + [-8.6060e-03, -1.8539e-03, -6.6833e-03, ..., 2.8381e-03, + -4.6692e-03, 1.7853e-03], + [ 9.9487e-03, -1.4038e-03, 5.2795e-03, ..., -7.4768e-03, + 7.4463e-03, -5.7983e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_norm.weight': tensor([0.1118, 0.0237, 0.4238, 0.0752, 0.1670, 0.0317, 0.1553, 0.1206, 0.5195, + 0.0452, 0.0664, 0.0161, 0.0518, 0.0674, 0.0640, 0.2041, 0.4590, 0.0344, + 0.7500, 0.0361, 0.0325, 0.1699, 0.0194, 0.0889, 0.0618, 0.4316, 0.5234, + 0.2734, 0.0767, 0.0349, 0.0150, 0.2324, 0.6719, 0.2539, 0.0474, 0.0271, + 0.0214, 0.0967, 0.0186, 0.0378, 0.0525, 0.1211, 0.0233, 0.0444, 0.2080, + 0.0498, 0.3555, 0.0262, 0.0557, 0.0208, 0.0262, 0.4023, 0.5312, 0.3789, + 0.1025, 0.5898, 0.5977, 0.0366, 0.1367, 0.0147, 0.0913, 0.4805, 0.0156, + 0.5273, 0.1689, 0.4160, 0.1011, 0.0498, 0.0732, 0.4297, 0.0282, 0.5508, + 0.0249, 0.0439, 0.1533, 0.0903, 0.0243, 0.1118, 0.6406, 0.0398, 0.0942, + 0.0879, 0.0292, 0.0461, 0.4219, 0.0679, 0.0698, 0.0850, 0.0175, 0.1553, + 0.0442, 0.1777, 0.0471, 0.0708, 0.0393, 0.1206, 0.0114, 0.3555, 0.5391, + 0.0498, 0.0552, 0.3242, 0.0339, 0.1309, 0.0398, 0.2285, 0.0156, 0.5742, + 0.5273, 0.0356, 0.0171, 0.7344, 0.0266, 0.0227, 0.4766, 0.2178, 0.3008, + 0.7188, 0.0364, 0.0981, 0.0801, 0.5625, 0.6055, 0.0198, 0.0188, 0.1914, + 0.0386, 0.0410], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.v_proj.weight': tensor([[ 0.0084, -0.0016, -0.0292, ..., 0.0041, 0.0038, -0.0220], + [ 0.0032, 0.0095, -0.0042, ..., 0.0006, -0.0060, -0.0199], + [ 0.0332, -0.0038, -0.0092, ..., -0.0049, -0.0175, -0.0386], + ..., + [-0.0118, 0.0019, 0.0160, ..., -0.0067, -0.0212, 0.0049], + [-0.0311, -0.0029, -0.0271, ..., -0.0032, 0.0128, -0.0242], + [-0.0053, 0.0071, 0.0076, ..., 0.0010, 0.0092, 0.0043]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0059, -0.0087, ..., -0.0060, 0.0027, 0.0042], + [-0.0173, -0.0145, -0.0193, ..., 0.0061, -0.0162, -0.0192], + [ 0.0070, -0.0029, -0.0039, ..., -0.0020, -0.0012, -0.0052], + ..., + [ 0.0082, -0.0060, 0.0193, ..., 0.0190, -0.0111, -0.0039], + [-0.0043, -0.0080, -0.0049, ..., 0.0244, -0.0148, 0.0232], + [-0.0089, -0.0129, 0.0269, ..., -0.0228, 0.0344, 0.0085]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0049, -0.0066, ..., 0.0266, -0.0278, 0.0076], + ..., + [-0.0217, -0.0410, 0.0488, ..., 0.0013, -0.0027, -0.0255], + [ 0.0015, 0.0066, 0.0024, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0057, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, 0.0027, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0033, 0.0293, ..., 0.0066, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0139, 0.0126, -0.0092], + ..., + [-0.0043, 0.0320, -0.0233, ..., -0.0067, 0.0016, 0.0188], + [ 0.0134, -0.0352, 0.0267, ..., -0.0305, -0.0021, 0.0037], + [-0.0205, 0.0140, -0.0019, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0019, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0057, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0016], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, -0.0013, 0.0015], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0022, ..., -0.0103, 0.0073, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.mlp.layer1.weight': tensor([[ 0.0108, 0.0060, -0.0140, ..., 0.0150, 0.0309, -0.0071], + [-0.0115, 0.0194, 0.0139, ..., -0.0099, -0.0183, 0.0038], + [-0.0079, 0.0284, -0.0168, ..., 0.0104, -0.0039, 0.0002], + ..., + [ 0.0049, 0.0046, 0.0630, ..., 0.0175, 0.0330, 0.0019], + [ 0.0021, -0.0042, -0.0120, ..., -0.0060, -0.0045, -0.0141], + [ 0.0013, -0.0130, -0.0064, ..., -0.0143, -0.0232, -0.0150]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.mlp.layer2.weight': tensor([[-0.0043, -0.0125, -0.0120, ..., -0.0203, 0.0067, -0.0003], + [-0.0059, 0.0171, 0.0068, ..., 0.0078, 0.0133, 0.0211], + [-0.0038, 0.0016, 0.0134, ..., -0.0035, 0.0135, 0.0177], + ..., + [-0.0152, 0.0171, 0.0059, ..., -0.0117, -0.0147, 0.0052], + [ 0.0100, -0.0058, 0.0141, ..., 0.0024, 0.0131, 0.0106], + [-0.0130, -0.0079, 0.0022, ..., 0.0167, 0.0105, 0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0444, -0.0437, -0.0427, ..., 0.0010, 0.0008, 0.0009], + [-0.0413, -0.0403, -0.0393, ..., 0.0027, 0.0028, 0.0027], + [-0.0396, -0.0386, -0.0374, ..., 0.0022, 0.0023, 0.0023], + ..., + [ 0.0498, 0.0481, 0.0469, ..., -0.0036, -0.0037, -0.0037], + [ 0.0398, 0.0386, 0.0371, ..., -0.0031, -0.0031, -0.0030], + [-0.0396, -0.0386, -0.0378, ..., 0.0023, 0.0024, 0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0062, 0.0070, 0.0070, ..., 0.0150, 0.0175, 0.0078], + [-0.0017, -0.0014, -0.0012, ..., -0.0032, -0.0031, -0.0013], + [ 0.0013, -0.0071, -0.0061, ..., 0.0034, 0.0036, -0.0070], + ..., + [-0.0325, -0.0776, -0.0645, ..., 0.0221, 0.0236, -0.0718], + [-0.0105, -0.0099, -0.0103, ..., 0.0039, 0.0040, -0.0096], + [ 0.0082, 0.0050, 0.0067, ..., 0.0162, 0.0187, 0.0062]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_cross_attn.1.weight': tensor([[-8.2520e-02, -8.0566e-02, -7.7637e-02, ..., -1.0586e-04, + -1.2064e-04, -1.1063e-04], + [ 3.3691e-02, 3.2471e-02, 3.1982e-02, ..., 1.9222e-06, + 2.7180e-05, 1.4877e-04], + [ 3.0884e-02, 2.9785e-02, 2.8809e-02, ..., 2.7657e-04, + 3.0136e-04, 2.0695e-04], + ..., + [ 5.0293e-02, 4.8828e-02, 4.8340e-02, ..., -1.7929e-04, + -1.7262e-04, -1.0538e-04], + [ 5.4932e-02, 5.3467e-02, 5.2002e-02, ..., -1.6785e-04, + -1.0729e-04, -1.5450e-04], + [-2.5635e-02, -2.4902e-02, -2.4292e-02, ..., -6.2943e-05, + 9.0122e-05, 1.1253e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0039, -0.0059, 0.0014, ..., 0.0275, 0.0325, 0.0023], + [ 0.0033, -0.0044, 0.0059, ..., -0.0010, 0.0016, 0.0042], + [ 0.0018, -0.0019, 0.0026, ..., -0.0084, -0.0086, -0.0021], + ..., + [-0.0198, 0.0078, 0.0280, ..., 0.0178, 0.0287, -0.0089], + [-0.0222, 0.0126, 0.0347, ..., -0.0042, -0.0026, -0.0033], + [ 0.0179, -0.0112, -0.0062, ..., -0.0096, -0.0111, 0.0176]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_mlp.1.weight': tensor([[-0.0771, -0.0752, -0.0737, ..., 0.0002, 0.0003, 0.0003], + [ 0.0282, 0.0273, 0.0267, ..., 0.0003, 0.0004, 0.0004], + [-0.0991, -0.0977, -0.0962, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0723, -0.0708, -0.0698, ..., 0.0002, 0.0003, 0.0002], + [-0.0540, -0.0527, -0.0518, ..., 0.0003, 0.0001, 0.0001], + [ 0.0703, 0.0693, 0.0679, ..., -0.0002, -0.0002, -0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_mlp.2.weight': tensor([[-0.0166, -0.0019, -0.0118, ..., -0.0186, -0.0221, 0.0200], + [-0.0018, -0.0050, 0.0067, ..., 0.0004, 0.0036, -0.0035], + [-0.0019, 0.0020, -0.0027, ..., -0.0019, -0.0019, 0.0019], + ..., + [-0.0486, -0.0170, -0.0557, ..., -0.0356, -0.0126, 0.0272], + [ 0.0071, -0.0099, 0.0264, ..., 0.0056, 0.0048, -0.0047], + [ 0.0138, 0.0110, 0.0107, ..., 0.0124, 0.0065, -0.0101]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.after_proj.weight': tensor([[ 0.0066, -0.0078, -0.0059, ..., -0.0055, 0.0026, -0.0152], + [ 0.0037, -0.0080, 0.0184, ..., -0.0098, -0.0061, -0.0095], + [ 0.0123, 0.0111, 0.0305, ..., 0.0052, 0.0048, -0.0076], + ..., + [ 0.0073, 0.0186, -0.0070, ..., 0.0038, -0.0143, -0.0076], + [-0.0108, -0.0059, -0.0055, ..., 0.0070, -0.0120, 0.0055], + [-0.0166, 0.0165, -0.0107, ..., -0.0049, 0.0009, 0.0073]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.after_proj.bias': tensor([ 0.0026, -0.0143, -0.0110, ..., 0.0014, -0.0015, 0.0062], + dtype=torch.bfloat16)} +[06-26 08:14:40|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:14:40|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 08:14:40|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 3 checkpoint. +[06-26 08:14:40|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:1015:compile_tokenizer_if_enabled] Tokenizer compilation disabled +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:146:__init__] Found 2 hint keys across all samples +[06-26 08:14:40|WARNING|cosmos_transfer2/inference.py:148:__init__] Loading the multicontrol model. Multicontrol inference is not strictly equal to single control +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/config.yaml +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['rainy_weather_lighting'] +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample rainy_weather_lighting +[06-26 08:14:40|DEBUG|cosmos_transfer2/inference.py:186:_generate_sample] InferenceArguments(name='rainy_weather_lighting' prompt_path=None prompt='The video shows an urban intersection captured from a vehicle-mounted camera during heavy rain. A flat overcast grey sky hangs low over the scene, the asphalt is dark, wet and mirror-like with rippling puddles and reflections of the surroundings, rain streaks the air and droplets bead on surfaces. The painted lane markings, stop lines and pedestrian crosswalks show through the glossy wet road. Multiple vehicles move through the junction with headlights on, low-rise buildings, sidewalks and street lamps line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, moody overcast lighting, detail.' negative_prompt='The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Bright sunlight, blue sky, dry road, hard crisp shadows.' seed=2025 guidance=3 video_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4') max_frames=None context_frame_index=None image_context_path=None num_conditional_frames=1 resolution='720' sigma_max=None num_video_frames_per_chunk=93 num_steps=35 show_control_condition=False show_input=False keep_input_resolution=True edge=EdgeConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4'), control_weight=1.0, mask_path=None, mask_prompt=None, preset_edge_threshold='medium') depth=None vis=BlurConfig(control_path=None, control_weight=0.2, mask_path=None, mask_prompt=None, preset_blur_strength='medium') seg=None guided_generation_mask=None guided_generation_step_threshold=25 guided_generation_foreground_labels=None) +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.json +[06-26 08:14:40|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:14:40|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None, 'vis': None, 'vis_mask': None, 'vis_mask_prompt': None} +[06-26 08:14:40|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:472:generate_img2world] Loading input video... +[06-26 08:14:40|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4 +[06-26 08:14:41|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:14:41|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:492:generate_img2world] Computing prompt text embeddings... +[06-26 08:14:41|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:501:generate_img2world] Computing negative prompt text embeddings... +[06-26 08:14:41|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:513:generate_img2world] Processing image context if available... +[06-26 08:14:41|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:330:read_and_process_image_context] No image context provided. +[06-26 08:14:41|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:525:generate_img2world] Loading control inputs... +[06-26 08:14:41|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4 +[06-26 08:14:41|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:14:42|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:703:read_and_process_control_input] No vis (blur) control input file found, will compute online.. +[06-26 08:14:42|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 1/2 +[06-26 08:14:48|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 584845 +[06-26 08:14:49|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 0 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:20:49|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 2/2 +[06-26 08:20:56|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 643774 +[06-26 08:20:57|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 1 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:26:39|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:736:generate_img2world] Average time per chunk: 356.68612208500053 +[06-26 08:26:39|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:26:40|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_edge.mp4 +[06-26 08:26:40|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:26:41|INFO|cosmos_transfer2/inference.py:311:_generate_sample] vis control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_vis.mp4 +[06-26 08:26:41|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:26:42|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:26:43|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.mp4 diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.json b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.json new file mode 100644 index 0000000000000000000000000000000000000000..916f5381fe903666f06794f6738d9aaa609bb181 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.json @@ -0,0 +1 @@ +{"name":"rainy_weather_lighting","prompt_path":null,"prompt":"The video shows an urban intersection captured from a vehicle-mounted camera during heavy rain. A flat overcast grey sky hangs low over the scene, the asphalt is dark, wet and mirror-like with rippling puddles and reflections of the surroundings, rain streaks the air and droplets bead on surfaces. The painted lane markings, stop lines and pedestrian crosswalks show through the glossy wet road. Multiple vehicles move through the junction with headlights on, low-rise buildings, sidewalks and street lamps line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, moody overcast lighting, detail.","negative_prompt":"The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Bright sunlight, blue sky, dry road, hard crisp shadows.","seed":2025,"guidance":3,"video_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4","max_frames":null,"context_frame_index":null,"image_context_path":null,"num_conditional_frames":1,"resolution":"720","sigma_max":null,"num_video_frames_per_chunk":93,"num_steps":35,"show_control_condition":false,"show_input":false,"keep_input_resolution":true,"edge":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4","control_weight":1.0,"mask_path":null,"mask_prompt":null,"preset_edge_threshold":"medium"},"depth":null,"vis":{"control_path":null,"control_weight":0.2,"mask_path":null,"mask_prompt":null,"preset_blur_strength":"medium"},"seg":null,"guided_generation_mask":null,"guided_generation_step_threshold":25,"guided_generation_foreground_labels":null} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.mp4 b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..323c67ce6e256e84cebe6864f9cfa8b2ca3c906e --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82253484c468eee39a59f4bad1a9c77729ba8c58ca1c21db53707f25a1753277 +size 1873274 diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.txt b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.txt new file mode 100644 index 0000000000000000000000000000000000000000..4987ef5fda9ebef26e8337960223fe1a39f70c28 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting.txt @@ -0,0 +1 @@ +The video shows an urban intersection captured from a vehicle-mounted camera during heavy rain. A flat overcast grey sky hangs low over the scene, the asphalt is dark, wet and mirror-like with rippling puddles and reflections of the surroundings, rain streaks the air and droplets bead on surfaces. The painted lane markings, stop lines and pedestrian crosswalks show through the glossy wet road. Multiple vehicles move through the junction with headlights on, low-rise buildings, sidewalks and street lamps line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, moody overcast lighting, detail. \ No newline at end of file diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_edge.mp4 b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..5cdebe53fd6e5d64099ea77a2b322d3c74554a2d --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d1a2bd621384e9ffdfe7aeacfdc3c277e04360f64aa63573dd720fda2ac14af +size 2253955 diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_vis.mp4 b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_vis.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..3954839e8e5b25fcae313f021c538aa5f31d799e --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/rainy_weather_lighting_control_vis.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9a9e148f1100e8757fd79f2d91e144f13d1ef9f647745a7550adf9926d0f418 +size 547205 diff --git a/multicontrol_cookbook_grid/rainy/weather/lighting/spec.json b/multicontrol_cookbook_grid/rainy/weather/lighting/spec.json new file mode 100644 index 0000000000000000000000000000000000000000..ff46fb711eb7d9ab2a8a2f8bad61588a45f31271 --- /dev/null +++ b/multicontrol_cookbook_grid/rainy/weather/lighting/spec.json @@ -0,0 +1,16 @@ +{ + "name": "rainy_weather_lighting", + "prompt": "The video shows an urban intersection captured from a vehicle-mounted camera during heavy rain. A flat overcast grey sky hangs low over the scene, the asphalt is dark, wet and mirror-like with rippling puddles and reflections of the surroundings, rain streaks the air and droplets bead on surfaces. The painted lane markings, stop lines and pedestrian crosswalks show through the glossy wet road. Multiple vehicles move through the junction with headlights on, low-rise buildings, sidewalks and street lamps line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, moody overcast lighting, detail.", + "video_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000_rainy/videos/rgb_front.mp4", + "guidance": 3, + "num_steps": 35, + "seed": 2025, + "edge": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/rainy_edge.mp4", + "control_weight": 1.0 + }, + "vis": { + "control_weight": 0.2 + }, + "negative_prompt": "The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Bright sunlight, blue sky, dry road, hard crisp shadows." +} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/Building/color/config.yaml b/multicontrol_cookbook_grid/sunny/Building/color/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9ce6429c128c2730a3d245a24c3c1dd969e9412b --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/Building/color/config.yaml @@ -0,0 +1,882 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + keys_to_skip_loading: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'False' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataloaders: + image_data: + ratio: '0' + video_data: + dataloader: + batch_size: '1' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '32' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control_and_image_context + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' + use_cache: 'False' + ratio: '1' + video_data_1: + dataloader: + batch_size: '6' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '32' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control_and_image_context + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + num_video_frames: '1' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' + use_cache: 'False' + ratio: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_control_conditioner +- ema: power +- tokenizer: wan2pt1_tokenizer +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: vid2vid_2B_control + name: vid2vid_2B_control_720p_t24_control_layer4_cr1pt1_embedding_rectified_flow_with_image_context_with_image_data + project: cosmos_transfer2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + base_load_from: null + conditional_frame_timestep: -1.0 + conditional_frames_probs: + 0: 0.4 + 1: 0.4 + 2: 0.2 + conditioner: + _target_: + control_input_depth: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth + output_key: control_input_depth + control_input_depth_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth_mask + output_key: control_input_depth_mask + control_input_edge: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge + output_key: control_input_edge + control_input_edge_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge_mask + output_key: control_input_edge_mask + control_input_inpaint: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint + output_key: control_input_inpaint + control_input_inpaint_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint_mask + output_key: control_input_inpaint_mask + control_input_seg: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg + output_key: control_input_seg + control_input_seg_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg_mask + output_key: control_input_seg_mask + control_input_vis: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis + output_key: control_input_vis + control_input_vis_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis_mask + output_key: control_input_vis_mask + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + reference_image_context: + _target_: + dropout_rate: '0.5' + input_key: + - images + - video + - image_context + num_token: '256' + output_key: null + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + copy_weight_strategy: first_n + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + hint_keys: edge + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + condition_strategy: spaced + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_image_context_dim: '1152' + extra_per_block_abs_pos_emb: 'False' + img_context_deep_proj: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + mlp_ratio: '4.0' + model_channels: '2048' + num_blocks: '28' + num_control_branches: '1' + num_heads: '16' + num_max_modalities: '8' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + separate_embedders: 'False' + share_q_in_i2v_cross_attn: 'False' + spatial_compression_factor: '8' + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_after_proj_for_multi_branch: 'True' + use_crossattn_projection: 'True' + use_cuda_graphs: 'False' + use_input_hint_block: 'False' + use_wan_fp32_strategy: 'True' + vace_block_every_n: '7' + vace_has_mask: 'False' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: uniform + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: false + use_lora: false + use_reference_image: true + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 8 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: '8.63e-05' + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '100000' + f_max: + - '0.5' + f_min: + - '0.2' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'True' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'False' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + frame_loss_log: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '100' + hit_thres: '300' + save_s3: 'True' + save_s3_every_log_n: '10' + load_base_model: + _target_: + config: null + trainer: null + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + compile_config: + recompile_limit: '8' + use_duck_shape: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '100000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/multicontrol_cookbook_grid/sunny/Building/color/console.log b/multicontrol_cookbook_grid/sunny/Building/color/console.log new file mode 100644 index 0000000000000000000000000000000000000000..b1113d7e420b7772fb6e5279db7c2378572b1ae5 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/Building/color/console.log @@ -0,0 +1,10 @@ +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:146:__init__] Found 1 hint keys across all samples +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color/config.yaml +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['sunny_Building_color'] +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample sunny_Building_color +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.json +[06-26 07:51:31|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None} +[06-26 08:02:03|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color_control_edge.mp4 +[06-26 08:02:03|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:02:04|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.mp4 diff --git a/multicontrol_cookbook_grid/sunny/Building/color/debug.log b/multicontrol_cookbook_grid/sunny/Building/color/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..7f44a4be2c961288fdda0d999a228aef10c9d873 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/Building/color/debug.log @@ -0,0 +1,969 @@ +[06-26 07:49:11|DEBUG|packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, training=True, smoke=False, verbose=False, experimental_checkpoints=False) +[06-26 07:49:12|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 07:49:12|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 07:49:23|CRITICAL|cosmos_transfer2/_src/interactive/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 07:49:23|CRITICAL|cosmos_transfer2/_src/predict2/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 07:49:23|DEBUG|cosmos_transfer2/inference.py:46:__init__] SetupArguments(output_dir=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color') model='edge' checkpoint_path='s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000' experiment='edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5' config_file='cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/config.py' context_parallel_size=1 disable_guardrails=True offload_guardrail_models=False keep_going=True profile=False benchmark=False compile_tokenizer= enable_parallel_tokenizer=False parallel_tokenizer_grid=(-1, -1))(['edge']) +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1... +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2... +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor... +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1_with_control... +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control... +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor_with_control... +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control_and_image_context... +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_with_control_input... +[06-26 07:49:24|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering hdmap_augmentor_for_local_datasets... +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package ('flash_attn_3_nv'_) was not found. +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version 0.21.0, expected at least 0.21.5.dev9. +[06-26 07:49:24|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_large_scale +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_small_overfit +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.experiment_list +[06-26 07:49:24|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.exp_large_scale +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.experiment_list +[06-26 07:49:24|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2.experiments +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot.cosmos_agibot_posttrain_example +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview.cosmos_multiview_example +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview +[06-26 07:49:24|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview.cosmos_singleview_example +[06-26 07:49:24|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 07:49:24|INFO|cosmos_transfer2/_src/imaginaire/utils/misc.py:151:set_random_seed] Using random seed 0. +[06-26 07:49:24|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 07:49:24|WARNING|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[06-26 07:49:25|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[06-26 07:49:25|INFO|cosmos_transfer2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/f176dc95b4a70f53ce01c4b302851595e7322b00/tokenizer.pth +[06-26 07:49:25|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.49 s +[06-26 07:51:25|INFO|cosmos_transfer2/_src/transfer2/networks/siglip2.py:65:get_siglip2_model_processor] +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #4-control_input_edge: + ReMapkey + input key: control_input_edge + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #5-control_input_vis: + ReMapkey + input key: control_input_vis + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #6-control_input_depth: + ReMapkey + input key: control_input_depth + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #7-control_input_seg: + ReMapkey + input key: control_input_seg + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #8-control_input_inpaint: + ReMapkey + input key: control_input_inpaint + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #9-control_input_edge_mask: + ReMapkey + input key: control_input_edge_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge_mask + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #10-control_input_vis_mask: + ReMapkey + input key: control_input_vis_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis_mask + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #11-control_input_depth_mask: + ReMapkey + input key: control_input_depth_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth_mask + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #12-control_input_seg_mask: + ReMapkey + input key: control_input_seg_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg_mask + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #13-control_input_inpaint_mask: + ReMapkey + input key: control_input_inpaint_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint_mask + Dtype: None +[06-26 07:51:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #14-reference_image_context: + SigLip2EmbImgContext + input key: ['images', 'video', 'image_context'] + Param count: 0 + Trainable: None + Dropout rate: 0.5 + Input key: ['images', 'video', 'image_context'] + Output key: ['img_context_emb'] +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.08 s +[06-26 07:51:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.84 s +[06-26 07:51:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 121.34 s +[06-26 07:51:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 121.34 s +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:569:freeze_base_model] +Freezing base model + +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:607:freeze_base_model] ✓ Unfroze img_context_proj +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 0 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 1 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 2 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 3 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 4 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 5 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 6 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 7 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 8 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 9 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 10 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 11 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 12 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 13 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 14 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 15 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 16 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 17 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 18 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 19 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 20 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 21 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 22 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 23 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 24 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 25 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 26 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 27 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 0 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 1 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 2 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 3 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:779:load_base_model] Done loading the base model checkpoint. +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 0-th block weight to control net's 0-th block +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 1-th block weight to control net's 1-th block +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 2-th block weight to control net's 2-th block +[06-26 07:51:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 3-th block weight to control net's 3-th block +[06-26 07:51:26|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[06-26 07:51:26|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[06-26 07:51:26|INFO|cosmos_transfer2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[06-26 07:51:26|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[06-26 07:51:26|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[06-26 07:51:27|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[06-26 07:51:27|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[06-26 07:51:29|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[06-26 07:51:29|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[06-26 07:51:29|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:76:__init__] MinimalV4LVGControlVaceDiT( + (x_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=72, out_features=2048, bias=False) + ) + ) + (pos_embedder): VideoRopePosition3DEmb() + (t_embedder): Sequential( + (0): Timesteps() + (1): TimestepEmbedding( + (linear_1): Linear(in_features=2048, out_features=2048, bias=False) + (activation): SiLU() + (linear_2): Linear(in_features=2048, out_features=6144, bias=False) + ) + ) + (blocks): ModuleList( + (0-27): 28 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlAwareDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + ) + ) + ) + (final_layer): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): FinalLayer( + (layer_norm): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (linear): Linear(in_features=2048, out_features=64, bias=False) + (adaln_modulation): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=4096, bias=False) + ) + ) + ) + ) + ) + (t_embedding_norm): RMSNorm() + (img_context_proj): Sequential( + (0): Linear(in_features=1152, out_features=2048, bias=True) + (1): GELU(approximate='none') + ) + (crossattn_proj): Sequential( + (0): Linear(in_features=100352, out_features=1024, bias=True) + (1): GELU(approximate='none') + ) + (control_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=520, out_features=2048, bias=False) + ) + ) + (control_blocks): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) +) +[06-26 07:51:30|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 125.27 s +[06-26 07:51:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 07:51:30|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 07:51:30|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:51:31|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +[06-26 07:51:31|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:271:load_model_state_dict_from_checkpoint] Checkpoint weights loaded from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt: None +[06-26 07:51:31|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:273:load_model_state_dict_from_checkpoint] Checkpoint weights loaded successfully (strict=True). +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:1015:compile_tokenizer_if_enabled] Tokenizer compilation disabled +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:146:__init__] Found 1 hint keys across all samples +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color/config.yaml +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['sunny_Building_color'] +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample sunny_Building_color +[06-26 07:51:31|DEBUG|cosmos_transfer2/inference.py:186:_generate_sample] InferenceArguments(name='sunny_Building_color' prompt_path=None prompt='The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. The streets are lined with ornate classical stone buildings featuring columns, cornices and tall arched windows, their facades warmly lit by the sun. A road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, multiple vehicles move through the junction, and the camera follows the moving ego vehicle smoothly. Photorealistic, soft shadows, high detail.' negative_prompt='The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Modern glass towers, plain concrete blocks, melted or warped facades, floating disconnected structures.' seed=2025 guidance=3 video_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4') max_frames=None context_frame_index=None image_context_path=None num_conditional_frames=1 resolution='720' sigma_max=None num_video_frames_per_chunk=93 num_steps=35 show_control_condition=False show_input=False keep_input_resolution=True edge=EdgeConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4'), control_weight=1.0, mask_path=None, mask_prompt=None, preset_edge_threshold='medium') depth=None vis=None seg=None guided_generation_mask=None guided_generation_step_threshold=25 guided_generation_foreground_labels=None) +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.json +[06-26 07:51:31|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 07:51:31|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None} +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:472:generate_img2world] Loading input video... +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4 +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:492:generate_img2world] Computing prompt text embeddings... +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:501:generate_img2world] Computing negative prompt text embeddings... +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:513:generate_img2world] Processing image context if available... +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:330:read_and_process_image_context] No image context provided. +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:525:generate_img2world] Loading control inputs... +[06-26 07:51:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4 +[06-26 07:51:32|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 07:51:32|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 1/2 +[06-26 07:51:32|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 584845 +[06-26 07:51:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 0 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 07:51:32|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:255:get_condition_with_negative_prompt] Getting condition with image context +[06-26 07:51:32|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 07:51:32|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:257:get_condition_with_negative_prompt] Getting uncondition with negative prompt without image context +[06-26 07:51:32|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 07:51:50|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 07:56:57|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 2/2 +[06-26 07:56:57|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 643774 +[06-26 07:56:57|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 1 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 07:56:57|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:255:get_condition_with_negative_prompt] Getting condition with image context +[06-26 07:56:57|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 07:56:57|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:257:get_condition_with_negative_prompt] Getting uncondition with negative prompt without image context +[06-26 07:56:57|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 07:56:59|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:02:02|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:736:generate_img2world] Average time per chunk: 313.7647706360003 +[06-26 08:02:02|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:02:03|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color_control_edge.mp4 +[06-26 08:02:03|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:02:03|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:02:04|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.mp4 diff --git a/multicontrol_cookbook_grid/sunny/Building/color/spec.json b/multicontrol_cookbook_grid/sunny/Building/color/spec.json new file mode 100644 index 0000000000000000000000000000000000000000..e7210a85e1d5081c9544d825812ace9fe51f1e49 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/Building/color/spec.json @@ -0,0 +1,13 @@ +{ + "name": "sunny_Building_color", + "prompt": "The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. The streets are lined with ornate classical stone buildings featuring columns, cornices and tall arched windows, their facades warmly lit by the sun. A road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, multiple vehicles move through the junction, and the camera follows the moving ego vehicle smoothly. Photorealistic, soft shadows, high detail.", + "video_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4", + "guidance": 3, + "num_steps": 35, + "seed": 2025, + "edge": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4", + "control_weight": 1.0 + }, + "negative_prompt": "The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Modern glass towers, plain concrete blocks, melted or warped facades, floating disconnected structures." +} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.json b/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.json new file mode 100644 index 0000000000000000000000000000000000000000..f746d3222a0ae82e6497c6cafda75413d9895fa0 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.json @@ -0,0 +1 @@ +{"name":"sunny_Building_color","prompt_path":null,"prompt":"The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. The streets are lined with ornate classical stone buildings featuring columns, cornices and tall arched windows, their facades warmly lit by the sun. A road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, multiple vehicles move through the junction, and the camera follows the moving ego vehicle smoothly. Photorealistic, soft shadows, high detail.","negative_prompt":"The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Modern glass towers, plain concrete blocks, melted or warped facades, floating disconnected structures.","seed":2025,"guidance":3,"video_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4","max_frames":null,"context_frame_index":null,"image_context_path":null,"num_conditional_frames":1,"resolution":"720","sigma_max":null,"num_video_frames_per_chunk":93,"num_steps":35,"show_control_condition":false,"show_input":false,"keep_input_resolution":true,"edge":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4","control_weight":1.0,"mask_path":null,"mask_prompt":null,"preset_edge_threshold":"medium"},"depth":null,"vis":null,"seg":null,"guided_generation_mask":null,"guided_generation_step_threshold":25,"guided_generation_foreground_labels":null} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.mp4 b/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..e9fc5118e5db9d0b76c19f9421a666a9fbf3d6c4 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:411c350a90edf73273a7985b3cb047e33bb6a88e800c48958e0a5cf0a2a9fa23 +size 1895086 diff --git a/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.txt b/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.txt new file mode 100644 index 0000000000000000000000000000000000000000..20a963b13c72a11374e0cef9d7e75383860a19e6 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color.txt @@ -0,0 +1 @@ +The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. The streets are lined with ornate classical stone buildings featuring columns, cornices and tall arched windows, their facades warmly lit by the sun. A road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, multiple vehicles move through the junction, and the camera follows the moving ego vehicle smoothly. Photorealistic, soft shadows, high detail. \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color_control_edge.mp4 b/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color_control_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..517a976ec0a5d8e4a484eb65a5ac5ad8308f733a --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/Building/color/sunny_Building_color_control_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57c1f383a1f61fc52a275e2a3d237bee50a45e7901dc709b401226496891cd0c +size 1406752 diff --git a/multicontrol_cookbook_grid/sunny/background/background/config.yaml b/multicontrol_cookbook_grid/sunny/background/background/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..61932c604f587593be0cf22b3130030c4b877da1 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/config.yaml @@ -0,0 +1,825 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + keys_to_skip_loading: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'False' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataset: + augmentor_name: video_basic_augmentor_v2_with_control + caption_type: t2w_qwen2p5_7b + control_input_type: edge_vis_depth_segcolor + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_control_conditioner +- ema: power +- tokenizer: wan2pt1_tokenizer +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: vid2vid_2B_control + name: vid2vid_2B_control_720p_t24_control_layer4_cr1pt1_embedding_rectified_flow + project: cosmos_transfer2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + base_load_from: null + conditional_frame_timestep: -1.0 + conditional_frames_probs: + 0: 0.4 + 1: 0.4 + 2: 0.2 + conditioner: + _target_: + control_input_depth: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth + output_key: control_input_depth + control_input_depth_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth_mask + output_key: control_input_depth_mask + control_input_edge: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge + output_key: control_input_edge + control_input_edge_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge_mask + output_key: control_input_edge_mask + control_input_inpaint: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint + output_key: control_input_inpaint + control_input_inpaint_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint_mask + output_key: control_input_inpaint_mask + control_input_seg: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg + output_key: control_input_seg + control_input_seg_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg_mask + output_key: control_input_seg_mask + control_input_vis: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis + output_key: control_input_vis + control_input_vis_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis_mask + output_key: control_input_vis_mask + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + copy_weight_strategy: first_n + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + hint_keys: edge_vis_depth_seg + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + condition_strategy: spaced + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_per_block_abs_pos_emb: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + mlp_ratio: '4.0' + model_channels: '2048' + num_blocks: '28' + num_control_branches: '4' + num_heads: '16' + num_max_modalities: '8' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + separate_embedders: 'False' + spatial_compression_factor: '8' + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_after_proj_for_multi_branch: 'False' + use_crossattn_projection: 'True' + use_cuda_graphs: 'False' + use_input_hint_block: 'False' + use_wan_fp32_strategy: 'True' + vace_block_every_n: '7' + vace_has_mask: 'False' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: reweighting + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: false + use_lora: false + use_reference_image: false + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 8 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: '8.63e-05' + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '100000' + f_max: + - '0.5' + f_min: + - '0.2' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'True' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'False' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + frame_loss_log: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '100' + hit_thres: '300' + save_s3: 'True' + save_s3_every_log_n: '10' + load_base_model: + _target_: + config: null + trainer: null + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + compile_config: + recompile_limit: '8' + use_duck_shape: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '100000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/multicontrol_cookbook_grid/sunny/background/background/console.log b/multicontrol_cookbook_grid/sunny/background/background/console.log new file mode 100644 index 0000000000000000000000000000000000000000..7dae0e4aead4bdd3564cebd8f79d4e8a73bfc136 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/console.log @@ -0,0 +1,13 @@ +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:146:__init__] Found 3 hint keys across all samples +[06-26 07:35:45|WARNING|cosmos_transfer2/inference.py:148:__init__] Loading the multicontrol model. Multicontrol inference is not strictly equal to single control +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/config.yaml +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['sunny_background_background'] +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample sunny_background_background +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.json +[06-26 07:35:45|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None, 'vis': None, 'vis_mask': None, 'vis_mask_prompt': None, 'seg': '/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/semseg_front.mp4', 'seg_mask': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4', 'seg_mask_prompt': None} +[06-26 07:49:05|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_edge.mp4 +[06-26 07:49:06|INFO|cosmos_transfer2/inference.py:311:_generate_sample] vis control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_vis.mp4 +[06-26 07:49:06|INFO|cosmos_transfer2/inference.py:311:_generate_sample] seg control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_seg.mp4 +[06-26 07:49:06|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 07:49:08|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.mp4 diff --git a/multicontrol_cookbook_grid/sunny/background/background/debug.log b/multicontrol_cookbook_grid/sunny/background/background/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..ccc3c2aeaa28d9f7fb883898030d385534128062 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/debug.log @@ -0,0 +1,33584 @@ +[06-26 07:35:21|DEBUG|packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, training=True, smoke=False, verbose=False, experimental_checkpoints=False) +[06-26 07:35:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 07:35:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 07:35:32|CRITICAL|cosmos_transfer2/_src/interactive/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 07:35:32|CRITICAL|cosmos_transfer2/_src/predict2/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 07:35:33|DEBUG|cosmos_transfer2/inference.py:46:__init__] SetupArguments(output_dir=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background') model='edge' checkpoint_path='s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000' experiment='edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5' config_file='cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/config.py' context_parallel_size=1 disable_guardrails=True offload_guardrail_models=False keep_going=True profile=False benchmark=False compile_tokenizer= enable_parallel_tokenizer=False parallel_tokenizer_grid=(-1, -1))(['edge', 'vis', 'seg']) +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1... +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2... +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor... +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1_with_control... +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control... +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor_with_control... +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control_and_image_context... +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_with_control_input... +[06-26 07:35:33|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering hdmap_augmentor_for_local_datasets... +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package ('flash_attn_3_nv'_) was not found. +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version 0.21.0, expected at least 0.21.5.dev9. +[06-26 07:35:33|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_large_scale +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_small_overfit +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.experiment_list +[06-26 07:35:33|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.exp_large_scale +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.experiment_list +[06-26 07:35:33|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2.experiments +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot.cosmos_agibot_posttrain_example +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview.cosmos_multiview_example +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview +[06-26 07:35:33|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview.cosmos_singleview_example +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 07:35:34|INFO|cosmos_transfer2/_src/imaginaire/utils/misc.py:151:set_random_seed] Using random seed 0. +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 07:35:34|WARNING|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/f176dc95b4a70f53ce01c4b302851595e7322b00/tokenizer.pth +[06-26 07:35:34|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.50 s +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #4-control_input_edge: + ReMapkey + input key: control_input_edge + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #5-control_input_vis: + ReMapkey + input key: control_input_vis + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #6-control_input_depth: + ReMapkey + input key: control_input_depth + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #7-control_input_seg: + ReMapkey + input key: control_input_seg + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #8-control_input_inpaint: + ReMapkey + input key: control_input_inpaint + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #9-control_input_edge_mask: + ReMapkey + input key: control_input_edge_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge_mask + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #10-control_input_vis_mask: + ReMapkey + input key: control_input_vis_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis_mask + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #11-control_input_depth_mask: + ReMapkey + input key: control_input_depth_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth_mask + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #12-control_input_seg_mask: + ReMapkey + input key: control_input_seg_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg_mask + Dtype: None +[06-26 07:35:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #13-control_input_inpaint_mask: + ReMapkey + input key: control_input_inpaint_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint_mask + Dtype: None +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:35:35|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.09 s +[06-26 07:35:35|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.80 s +[06-26 07:35:35|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 0.81 s +[06-26 07:35:35|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 0.81 s +[06-26 07:35:35|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:569:freeze_base_model] +Freezing base model + +[06-26 07:35:35|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:779:load_base_model] Done loading the base model checkpoint. +[06-26 07:35:35|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 0-th block weight to control net's 0-th block +[06-26 07:35:35|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 1-th block weight to control net's 1-th block +[06-26 07:35:35|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 2-th block weight to control net's 2-th block +[06-26 07:35:35|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 3-th block weight to control net's 3-th block +[06-26 07:35:35|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[06-26 07:35:35|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[06-26 07:35:36|INFO|cosmos_transfer2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[06-26 07:35:36|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[06-26 07:35:36|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[06-26 07:35:36|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[06-26 07:35:36|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[06-26 07:35:39|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[06-26 07:35:39|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[06-26 07:35:39|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:76:__init__] MinimalV4LVGControlVaceDiT( + (x_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=72, out_features=2048, bias=False) + ) + ) + (pos_embedder): VideoRopePosition3DEmb() + (t_embedder): Sequential( + (0): Timesteps() + (1): TimestepEmbedding( + (linear_1): Linear(in_features=2048, out_features=2048, bias=False) + (activation): SiLU() + (linear_2): Linear(in_features=2048, out_features=6144, bias=False) + ) + ) + (blocks): ModuleList( + (0-27): 28 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlAwareDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + ) + ) + ) + (final_layer): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): FinalLayer( + (layer_norm): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (linear): Linear(in_features=2048, out_features=64, bias=False) + (adaln_modulation): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=4096, bias=False) + ) + ) + ) + ) + ) + ) + ) + ) + (t_embedding_norm): RMSNorm() + (crossattn_proj): Sequential( + (0): Linear(in_features=100352, out_features=1024, bias=True) + (1): GELU(approximate='none') + ) + (control_embedder): ModuleList( + (0-3): 4 x PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=520, out_features=2048, bias=False) + ) + ) + ) + (control_blocks_0): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_1): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_2): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_3): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) +) +[06-26 07:35:39|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 5.16 s +[06-26 07:35:39|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 07:35:39|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 07:35:40|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:40|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'control_blocks.0.self_attn.q_proj.weight', 'control_blocks.0.self_attn.q_norm.weight', 'control_blocks.0.self_attn.k_proj.weight', 'control_blocks.0.self_attn.k_norm.weight', 'control_blocks.0.self_attn.v_proj.weight', 'control_blocks.0.self_attn.output_proj.weight', 'control_blocks.0.cross_attn.q_proj.weight', 'control_blocks.0.cross_attn.q_norm.weight', 'control_blocks.0.cross_attn.k_proj.weight', 'control_blocks.0.cross_attn.k_norm.weight', 'control_blocks.0.cross_attn.v_proj.weight', 'control_blocks.0.cross_attn.output_proj.weight', 'control_blocks.0.cross_attn.k_img.weight', 'control_blocks.0.cross_attn.v_img.weight', 'control_blocks.0.cross_attn.q_img.weight', 'control_blocks.0.cross_attn.q_img_norm.weight', 'control_blocks.0.cross_attn.k_img_norm.weight', 'control_blocks.0.mlp.layer1.weight', 'control_blocks.0.mlp.layer2.weight', 'control_blocks.0.adaln_modulation_self_attn.1.weight', 'control_blocks.0.adaln_modulation_self_attn.2.weight', 'control_blocks.0.adaln_modulation_cross_attn.1.weight', 'control_blocks.0.adaln_modulation_cross_attn.2.weight', 'control_blocks.0.adaln_modulation_mlp.1.weight', 'control_blocks.0.adaln_modulation_mlp.2.weight', 'control_blocks.0.before_proj.weight', 'control_blocks.0.before_proj.bias', 'control_blocks.0.after_proj.weight', 'control_blocks.0.after_proj.bias', 'control_blocks.1.self_attn.q_proj.weight', 'control_blocks.1.self_attn.q_norm.weight', 'control_blocks.1.self_attn.k_proj.weight', 'control_blocks.1.self_attn.k_norm.weight', 'control_blocks.1.self_attn.v_proj.weight', 'control_blocks.1.self_attn.output_proj.weight', 'control_blocks.1.cross_attn.q_proj.weight', 'control_blocks.1.cross_attn.q_norm.weight', 'control_blocks.1.cross_attn.k_proj.weight', 'control_blocks.1.cross_attn.k_norm.weight', 'control_blocks.1.cross_attn.v_proj.weight', 'control_blocks.1.cross_attn.output_proj.weight', 'control_blocks.1.cross_attn.k_img.weight', 'control_blocks.1.cross_attn.v_img.weight', 'control_blocks.1.cross_attn.q_img.weight', 'control_blocks.1.cross_attn.q_img_norm.weight', 'control_blocks.1.cross_attn.k_img_norm.weight', 'control_blocks.1.mlp.layer1.weight', 'control_blocks.1.mlp.layer2.weight', 'control_blocks.1.adaln_modulation_self_attn.1.weight', 'control_blocks.1.adaln_modulation_self_attn.2.weight', 'control_blocks.1.adaln_modulation_cross_attn.1.weight', 'control_blocks.1.adaln_modulation_cross_attn.2.weight', 'control_blocks.1.adaln_modulation_mlp.1.weight', 'control_blocks.1.adaln_modulation_mlp.2.weight', 'control_blocks.1.after_proj.weight', 'control_blocks.1.after_proj.bias', 'control_blocks.2.self_attn.q_proj.weight', 'control_blocks.2.self_attn.q_norm.weight', 'control_blocks.2.self_attn.k_proj.weight', 'control_blocks.2.self_attn.k_norm.weight', 'control_blocks.2.self_attn.v_proj.weight', 'control_blocks.2.self_attn.output_proj.weight', 'control_blocks.2.cross_attn.q_proj.weight', 'control_blocks.2.cross_attn.q_norm.weight', 'control_blocks.2.cross_attn.k_proj.weight', 'control_blocks.2.cross_attn.k_norm.weight', 'control_blocks.2.cross_attn.v_proj.weight', 'control_blocks.2.cross_attn.output_proj.weight', 'control_blocks.2.cross_attn.k_img.weight', 'control_blocks.2.cross_attn.v_img.weight', 'control_blocks.2.cross_attn.q_img.weight', 'control_blocks.2.cross_attn.q_img_norm.weight', 'control_blocks.2.cross_attn.k_img_norm.weight', 'control_blocks.2.mlp.layer1.weight', 'control_blocks.2.mlp.layer2.weight', 'control_blocks.2.adaln_modulation_self_attn.1.weight', 'control_blocks.2.adaln_modulation_self_attn.2.weight', 'control_blocks.2.adaln_modulation_cross_attn.1.weight', 'control_blocks.2.adaln_modulation_cross_attn.2.weight', 'control_blocks.2.adaln_modulation_mlp.1.weight', 'control_blocks.2.adaln_modulation_mlp.2.weight', 'control_blocks.2.after_proj.weight', 'control_blocks.2.after_proj.bias', 'control_blocks.3.self_attn.q_proj.weight', 'control_blocks.3.self_attn.q_norm.weight', 'control_blocks.3.self_attn.k_proj.weight', 'control_blocks.3.self_attn.k_norm.weight', 'control_blocks.3.self_attn.v_proj.weight', 'control_blocks.3.self_attn.output_proj.weight', 'control_blocks.3.cross_attn.q_proj.weight', 'control_blocks.3.cross_attn.q_norm.weight', 'control_blocks.3.cross_attn.k_proj.weight', 'control_blocks.3.cross_attn.k_norm.weight', 'control_blocks.3.cross_attn.v_proj.weight', 'control_blocks.3.cross_attn.output_proj.weight', 'control_blocks.3.cross_attn.k_img.weight', 'control_blocks.3.cross_attn.v_img.weight', 'control_blocks.3.cross_attn.q_img.weight', 'control_blocks.3.cross_attn.q_img_norm.weight', 'control_blocks.3.cross_attn.k_img_norm.weight', 'control_blocks.3.mlp.layer1.weight', 'control_blocks.3.mlp.layer2.weight', 'control_blocks.3.adaln_modulation_self_attn.1.weight', 'control_blocks.3.adaln_modulation_self_attn.2.weight', 'control_blocks.3.adaln_modulation_cross_attn.1.weight', 'control_blocks.3.adaln_modulation_cross_attn.2.weight', 'control_blocks.3.adaln_modulation_mlp.1.weight', 'control_blocks.3.adaln_modulation_mlp.2.weight', 'control_blocks.3.after_proj.weight', 'control_blocks.3.after_proj.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_embedder.proj.1.weight'], incorrect_shapes=[]) +[06-26 07:35:40|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:271:load_model_state_dict_from_checkpoint] Checkpoint weights loaded from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt: None +[06-26 07:35:40|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:273:load_model_state_dict_from_checkpoint] Checkpoint weights loaded successfully (strict=True). +[06-26 07:35:40|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 07:35:40|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 0 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 07:35:41|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0148, 0.0025, 0.0065], + [-0.0004, 0.0057, -0.0327, ..., -0.0078, 0.0021, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0136, -0.0035, 0.0283], + ..., + [-0.0028, 0.0125, 0.0248, ..., 0.0437, -0.0069, -0.0192], + [ 0.0277, 0.0146, 0.0031, ..., -0.0028, -0.0090, 0.0146], + [ 0.0437, 0.0366, 0.0082, ..., 0.0029, -0.0126, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0044, -0.0310, 0.0300, ..., -0.0064, -0.0045, -0.0339], + [-0.0103, -0.0047, 0.0114, ..., 0.0052, 0.0092, 0.0520], + [-0.0061, 0.0131, 0.0398, ..., -0.0096, -0.0179, -0.0056], + ..., + [-0.0048, 0.0175, -0.0101, ..., 0.0225, 0.0043, 0.0042], + [-0.0209, -0.0081, 0.0089, ..., -0.0308, 0.0108, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0055, 0.0354, -0.0082, ..., 0.0120, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0013, ..., 0.0376, 0.0532, 0.0036], + [ 0.0175, -0.0347, 0.0013, ..., -0.0066, 0.0053, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0154, 0.0262, -0.0112], + [-0.0006, 0.0244, 0.0356, ..., -0.0036, 0.0170, 0.0118], + [ 0.0486, 0.0175, 0.0042, ..., 0.0036, -0.0026, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0114, ..., -0.0070, -0.0337, 0.0077], + [ 0.0491, 0.0413, 0.0251, ..., -0.0030, 0.0393, 0.0042], + [-0.0162, -0.0019, 0.0140, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0038, 0.0261, 0.0039, ..., -0.0040, 0.0347, 0.0009], + [-0.0254, 0.0063, 0.0019, ..., 0.0093, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0121, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0035, -0.0270, -0.0166, ..., 0.0115, -0.0298, 0.0532], + [ 0.0017, 0.0040, -0.0131, ..., -0.0161, 0.0146, 0.0308], + [ 0.0146, 0.0087, 0.0143, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0064, -0.0630, -0.0048, ..., 0.0074, -0.0183, 0.0217], + [ 0.0023, -0.0035, 0.0258, ..., -0.0486, -0.0349, -0.0059], + [ 0.0182, -0.0076, 0.0197, ..., 0.0142, -0.0226, -0.0212]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0051, -0.0012, -0.0011, ..., -0.0005, -0.0018, -0.0120], + [-0.0153, 0.0255, 0.0432, ..., -0.0045, 0.0049, -0.0069], + [-0.0154, -0.0206, -0.0245, ..., -0.0119, -0.0325, -0.0013], + ..., + [-0.0034, 0.0039, 0.0417, ..., 0.0280, 0.0036, -0.0047], + [-0.0054, 0.0041, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0055, -0.0024, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 1.5503e-02, -6.5613e-03, -4.8096e-02, ..., 4.4922e-02, + 2.1240e-02, 1.2756e-02], + [ 1.5381e-02, -2.2705e-02, 2.1973e-02, ..., -2.2095e-02, + 3.5858e-03, 6.9885e-03], + [-1.5259e-02, -1.0071e-02, 4.2969e-02, ..., -9.5825e-03, + 3.2715e-02, -1.1169e-02], + ..., + [ 3.1006e-02, -3.7598e-02, -4.3457e-02, ..., -2.2461e-02, + 2.5513e-02, 1.3428e-02], + [-3.8862e-05, 1.6602e-02, -1.6174e-03, ..., -4.0283e-03, + 6.7139e-03, -9.6436e-03], + [-1.0864e-02, -3.4637e-03, 1.1780e-02, ..., 4.6997e-03, + -3.7994e-03, -7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[ 0.0049, -0.0386, -0.0112, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0079, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0127, 0.0032, -0.0114, ..., 0.0214, 0.0110, -0.0045], + ..., + [-0.0155, -0.0172, -0.0136, ..., 0.0095, 0.0098, 0.0184], + [-0.0151, -0.0356, 0.0200, ..., -0.0060, -0.0112, 0.0332], + [-0.0396, -0.0120, -0.0089, ..., -0.0284, -0.0140, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 2.7710e-02, 3.1250e-02, 1.3306e-02, ..., -1.9531e-02, + 1.1414e-02, -1.4954e-02], + [ 1.9165e-02, 9.2773e-03, -7.5378e-03, ..., -1.7853e-03, + -1.2695e-02, -1.9775e-02], + [-2.5024e-02, 1.1108e-02, -1.3733e-02, ..., 1.0132e-02, + 1.7212e-02, 7.0496e-03], + ..., + [ 4.6631e-02, 6.4392e-03, -1.5869e-03, ..., -4.0283e-03, + 2.3926e-02, 2.5330e-03], + [ 5.4688e-02, -1.8677e-02, -1.1108e-02, ..., -3.7537e-03, + 3.1128e-02, -5.4626e-03], + [ 8.7738e-05, 1.0742e-02, -9.6436e-03, ..., 2.0020e-02, + 1.6968e-02, 1.4038e-02]], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[-5.6028e-05, 2.3651e-03, 2.8564e-02, ..., -1.0376e-02, + 4.5471e-03, -1.8188e-02], + [ 6.2866e-03, 1.3428e-02, -3.1494e-02, ..., -4.5654e-02, + -5.0049e-03, 1.0376e-02], + [ 4.3945e-03, 2.6123e-02, 2.3926e-02, ..., 5.0049e-03, + 6.6833e-03, 4.6143e-02], + ..., + [ 1.6968e-02, -2.5391e-02, -4.3945e-02, ..., 1.5259e-02, + 4.9133e-03, -7.4768e-03], + [ 5.7678e-03, -3.7109e-02, 3.5645e-02, ..., -8.3008e-03, + 4.1016e-02, -1.7822e-02], + [ 1.7822e-02, -1.4648e-02, 5.6152e-02, ..., -2.9053e-02, + -2.6733e-02, 2.9663e-02]], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0067, 0.0208, 0.0278, ..., 0.0133, 0.0156, 0.0074], + [-0.0063, 0.0040, 0.0014, ..., 0.0135, 0.0040, 0.0045], + [ 0.0058, 0.0041, 0.0243, ..., -0.0123, -0.0049, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0064, ..., 0.0250, -0.0107, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., 0.0048, -0.0295, 0.0400], + [-0.0116, 0.0138, 0.0305, ..., -0.0103, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0054, 0.0381, -0.0239, ..., 0.0219, 0.0483, 0.0339], + [-0.0143, -0.0033, 0.0288, ..., 0.0135, 0.0583, 0.0049], + [-0.0280, 0.0154, -0.0386, ..., -0.0444, -0.0092, -0.0026], + ..., + [-0.0038, 0.0157, 0.0217, ..., 0.0048, 0.0114, 0.0154], + [ 0.0081, 0.0195, -0.0283, ..., -0.0060, -0.0021, -0.0342], + [-0.0166, -0.0195, 0.0027, ..., -0.0066, 0.0165, 0.0130]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0138, -0.0146, 0.0459, ..., -0.0028, -0.0320, 0.0277], + [-0.0026, 0.0068, 0.0270, ..., -0.0220, -0.0183, -0.0271], + [ 0.0221, 0.0219, -0.0201, ..., 0.0060, 0.0064, -0.0031], + ..., + [-0.0258, 0.0016, 0.0215, ..., 0.0038, 0.0237, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0055], + [-0.0006, -0.0080, -0.0120, ..., -0.0334, 0.0047, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0147, -0.0454, ..., -0.0132, -0.0125, 0.0400], + [ 0.0295, -0.0187, -0.0110, ..., -0.0283, 0.0047, -0.0186], + [-0.0007, 0.0547, 0.0024, ..., 0.0116, 0.0400, -0.0214], + ..., + [ 0.0280, -0.0058, -0.0266, ..., -0.0084, -0.0287, 0.0150], + [ 0.0090, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0095, 0.0102, ..., -0.0070, 0.0703, -0.0043]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0205, 0.0087, -0.0315, ..., 0.0081, 0.0254, 0.0359], + [ 0.0097, -0.0032, 0.0038, ..., -0.0120, -0.0233, -0.0048], + [ 0.0051, 0.0040, -0.0121, ..., -0.0117, -0.0070, 0.0086], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0040], + [ 0.0339, -0.0064, 0.0223, ..., 0.0410, -0.0017, 0.0136], + [-0.0046, -0.0439, -0.0486, ..., 0.0145, -0.0130, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0046, 0.0139, -0.0030, ..., -0.0081, -0.0033, -0.0229], + [-0.0289, 0.0231, 0.0077, ..., -0.0135, -0.0024, 0.0222], + [ 0.0083, 0.0295, -0.0189, ..., -0.0168, 0.0311, -0.0261], + ..., + [ 0.0233, -0.0045, -0.0036, ..., 0.0032, 0.0278, 0.0049], + [-0.0106, -0.0038, 0.0140, ..., -0.0175, 0.0444, 0.0425], + [ 0.0069, -0.0078, 0.0102, ..., 0.0066, -0.0344, 0.0229]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[-0.0050, 0.0128, 0.0190, ..., -0.0093, -0.0198, 0.0149], + [-0.0581, 0.0294, -0.0047, ..., -0.0240, 0.0043, 0.0300], + [ 0.0085, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0035, -0.0289, 0.0537, ..., 0.0058, -0.0074, -0.0099], + [-0.0029, -0.0069, 0.0210, ..., -0.0021, 0.0091, 0.0186], + [-0.0086, -0.0349, -0.0173, ..., -0.0400, 0.0052, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0177, -0.0261, -0.0112, ..., 0.0110, -0.0130, -0.0405], + [ 0.0058, -0.0086, 0.0266, ..., -0.0400, 0.0073, 0.0391], + [-0.0013, 0.0359, 0.0284, ..., 0.0286, -0.0234, 0.0217], + ..., + [ 0.0017, 0.0123, 0.0132, ..., -0.0138, 0.0032, 0.0096], + [ 0.0299, -0.0167, -0.0022, ..., -0.0063, 0.0043, -0.0206], + [ 0.0074, -0.0076, -0.0043, ..., -0.0047, -0.0067, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, -0.0032, 0.0121, ..., -0.0075, -0.0208, -0.0058], + [-0.0092, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0229], + [ 0.0132, -0.0278, 0.0115, ..., 0.0396, -0.0082, -0.0074], + ..., + [-0.0015, 0.0018, -0.0309, ..., 0.0344, 0.0073, -0.0126], + [-0.0236, -0.0102, -0.0056, ..., 0.0332, -0.0117, -0.0171], + [ 0.0231, 0.0405, 0.0103, ..., -0.0278, -0.0139, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[-0.0046, 0.0166, 0.0078, ..., 0.0273, -0.0352, 0.0153], + [-0.0510, -0.0070, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0077, ..., -0.0388, -0.0047, 0.0398], + ..., + [-0.0160, -0.0195, -0.0051, ..., -0.0040, -0.0055, -0.0035], + [-0.0187, -0.0040, 0.0097, ..., 0.0287, -0.0122, -0.0099], + [-0.0181, -0.0013, 0.0247, ..., -0.0145, 0.0072, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0078, 0.0092, 0.0140, ..., -0.0121, 0.0045, -0.0253], + [-0.0164, 0.0330, -0.0171, ..., -0.0033, -0.0403, -0.0111], + [ 0.0085, 0.0347, 0.0275, ..., 0.0151, 0.0069, 0.0080], + ..., + [ 0.0275, -0.0168, -0.0106, ..., -0.0099, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0056, ..., -0.0023, -0.0022, -0.0243], + [-0.0060, 0.0581, 0.0061, ..., 0.0148, 0.0239, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0121, -0.0410, 0.0128, ..., -0.0254, -0.0036, -0.0173], + [ 0.0476, 0.0315, 0.0141, ..., -0.0154, -0.0073, 0.0048], + [-0.0092, -0.0270, 0.0092, ..., 0.0332, 0.0150, -0.0073], + ..., + [-0.0076, 0.0039, 0.0162, ..., -0.0177, 0.0171, -0.0133], + [ 0.0044, -0.0187, -0.0051, ..., 0.0248, -0.0145, -0.0073], + [-0.0143, 0.0034, 0.0149, ..., -0.0047, -0.0006, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0049, 0.0097, ..., 0.0255, -0.0253, -0.0136], + [-0.0038, -0.0141, -0.0160, ..., -0.0109, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0117, 0.0056, 0.0322, ..., -0.0183, -0.0041, -0.0062], + [ 0.0045, -0.0054, 0.0164, ..., 0.0036, 0.0255, 0.0024], + [ 0.0270, 0.0084, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0077, ..., -0.0532, 0.0106, 0.0036], + [-0.0240, -0.0215, -0.0104, ..., -0.0063, 0.0097, -0.0168], + [ 0.0075, 0.0064, -0.0055, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0059, 0.0172, -0.0106, ..., -0.0300, -0.0005, 0.0435], + [ 0.0053, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0061], + [-0.0167, 0.0128, 0.0101, ..., -0.0247, 0.0119, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0065, -0.0079, 0.0317, ..., 0.0200, -0.0105, -0.0376], + [ 0.0356, 0.0378, 0.0129, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., 0.0076, 0.0065, -0.0322], + ..., + [ 0.0065, 0.0107, -0.0029, ..., 0.0366, -0.0206, 0.0374], + [-0.0034, -0.0079, -0.0044, ..., -0.0095, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0048, ..., -0.0166, -0.0042, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0078, -0.0096, -0.0344, ..., -0.0003, 0.0347, -0.0181], + [ 0.0267, -0.0366, -0.0089, ..., 0.0217, 0.0349, -0.0070], + [-0.0098, -0.0286, 0.0062, ..., 0.0060, -0.0282, 0.0047], + ..., + [-0.0124, -0.0039, 0.0065, ..., -0.0084, 0.0165, 0.0074], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0134], + [-0.0140, 0.0049, -0.0170, ..., -0.0126, 0.0060, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0033, 0.0051, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0120, 0.0374, ..., -0.0153, -0.0188, -0.0137], + ..., + [ 0.0262, 0.0052, -0.0065, ..., 0.0007, 0.0038, -0.0165], + [-0.0222, -0.0155, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0125, -0.0148, 0.0038, ..., 0.0018, -0.0234, 0.0017]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0102], + [ 0.0249, 0.0205, 0.0054, ..., -0.0046, -0.0093, -0.0106], + [-0.0220, -0.0004, -0.0037, ..., 0.0306, 0.0381, -0.0044], + ..., + [ 0.0200, -0.0050, 0.0170, ..., 0.0116, 0.0320, -0.0157], + [ 0.0349, 0.0119, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0049, 0.0256, 0.0045, ..., 0.0205, 0.0002, -0.0105]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0227, 0.0065, 0.0199, ..., -0.0057, 0.0276, -0.0312], + [ 0.0178, -0.0059, 0.0103, ..., 0.0112, -0.0398, 0.0293], + [ 0.0430, -0.0130, 0.0072, ..., 0.0040, -0.0069, -0.0029], + ..., + [ 0.0192, -0.0112, -0.0112, ..., 0.0027, 0.0330, 0.0137], + [-0.0125, -0.0542, -0.0205, ..., 0.0044, -0.0159, 0.0320], + [ 0.0303, 0.0047, -0.0425, ..., 0.0032, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0060, -0.0315, ..., 0.0525, 0.0041, -0.0349], + [-0.0036, -0.0026, -0.0156, ..., 0.0048, -0.0240, 0.0305], + [-0.0256, 0.0064, 0.0065, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0065, 0.0231, 0.0209, ..., 0.0076, -0.0041, 0.0461], + [-0.0063, -0.0079, -0.0019, ..., -0.0217, 0.0347, -0.0040], + [ 0.0050, 0.0234, 0.0031, ..., 0.0261, -0.0052, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0052, 0.0046], + [-0.0183, -0.0211, -0.0052, ..., -0.0149, 0.0139, -0.0162], + [-0.0078, -0.0160, 0.0079, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0096, 0.0073, ..., 0.0171, 0.0036, -0.0033], + [-0.0144, 0.0260, -0.0087, ..., 0.0041, 0.0106, -0.0166], + [ 0.0098, -0.0140, -0.0014, ..., 0.0057, 0.0054, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0078, 0.0134, -0.0064, ..., -0.0278, 0.0240, -0.0123], + [-0.0069, 0.0188, -0.0160, ..., -0.0317, -0.0064, 0.0181], + [-0.0042, 0.0310, -0.0085, ..., -0.0226, 0.0017, 0.0334], + ..., + [-0.0081, 0.0186, -0.0073, ..., 0.0228, 0.0047, 0.0064], + [ 0.0271, 0.0118, 0.0354, ..., -0.0075, 0.0198, 0.0150], + [ 0.0282, 0.0171, 0.0317, ..., -0.0109, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-1.1475e-02, -2.6489e-02, -1.0986e-02, ..., 4.8523e-03, + 3.7231e-03, 1.3428e-02], + [-3.3203e-02, -4.8523e-03, 3.0640e-02, ..., 3.4668e-02, + -3.6133e-02, 8.4839e-03], + [-5.8651e-05, -1.9043e-02, -3.6774e-03, ..., 1.6602e-02, + -1.3550e-02, 1.3733e-02], + ..., + [-2.0020e-02, -2.9144e-03, 2.4414e-04, ..., -3.3936e-02, + 3.6133e-02, 3.0518e-03], + [ 6.6528e-03, -2.5024e-02, 1.5564e-02, ..., 2.2430e-03, + -1.6235e-02, 2.1973e-02], + [-7.4463e-03, 3.8452e-03, 2.0508e-02, ..., -6.5002e-03, + -8.3008e-03, 4.3457e-02]], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0047, 0.0240, -0.0217, ..., 0.0172, 0.0046, 0.0114], + [-0.0386, 0.0106, 0.0078, ..., -0.0060, 0.0322, 0.0086], + [-0.0427, 0.0011, -0.0045, ..., 0.0076, 0.0112, -0.0137], + ..., + [-0.0087, -0.0037, -0.0264, ..., -0.0586, -0.0630, -0.0131], + [-0.0006, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0107, -0.0610, 0.0162, ..., -0.0291, -0.0166, 0.0029]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0117, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0112], + [ 0.0111, -0.0098, -0.0327, ..., 0.0447, 0.0076, 0.0059], + [ 0.0042, -0.0175, 0.0076, ..., -0.0069, -0.0037, -0.0060], + ..., + [ 0.0295, 0.0042, -0.0102, ..., 0.0099, -0.0160, -0.0286], + [ 0.0082, 0.0253, -0.0177, ..., 0.0095, 0.0064, 0.0078], + [-0.0132, -0.0048, 0.0139, ..., -0.0125, 0.0277, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0129, 0.0040, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0050, 0.0234, 0.0069, ..., 0.0464, 0.0056, 0.0309], + ..., + [-0.0118, -0.0175, 0.0327, ..., 0.0132, 0.0051, 0.0141], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, 0.0040, -0.0444, ..., 0.0060, -0.0283, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0081, -0.0100, 0.0146, ..., -0.0742, -0.0074, 0.0161], + [-0.0064, -0.0132, -0.0071, ..., 0.0053, 0.0042, 0.0238], + [ 0.0189, 0.0221, -0.0087, ..., 0.0033, 0.0408, 0.0064], + ..., + [ 0.0047, 0.0247, 0.0396, ..., 0.0032, -0.0304, -0.0183], + [ 0.0299, -0.0113, -0.0417, ..., -0.0052, -0.0376, 0.0032], + [-0.0210, 0.0168, -0.0039, ..., 0.0052, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0056, 0.0242, -0.0079, ..., -0.0245, 0.0050, -0.0057], + [ 0.0190, -0.0123, 0.0332, ..., -0.0143, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0053, ..., 0.0056, -0.0091, 0.0182], + ..., + [-0.0430, -0.0403, -0.0086, ..., 0.0134, -0.0140, 0.0060], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0089, 0.0378], + [ 0.0364, 0.0486, -0.0120, ..., -0.0060, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0094, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0078, ..., 0.0101, -0.0223, -0.0052], + ..., + [ 0.0192, 0.0292, -0.0022, ..., 0.0176, -0.0197, 0.0076], + [ 0.0265, 0.0064, 0.0057, ..., 0.0286, 0.0046, -0.0267], + [-0.0184, -0.0088, -0.0045, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0083, -0.0159, 0.0082, ..., 0.0198, 0.0044, 0.0277], + [ 0.0172, 0.0037, -0.0053, ..., 0.0054, -0.0044, 0.0074], + [-0.0019, 0.0184, 0.0113, ..., 0.0099, -0.0054, -0.0077], + ..., + [ 0.0061, -0.0111, -0.0028, ..., -0.0095, -0.0038, 0.0123], + [-0.0427, -0.0391, -0.0193, ..., 0.0096, -0.0238, -0.0078], + [ 0.0040, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0217, -0.0226, -0.0349, ..., -0.0079, -0.0161, 0.0127], + [-0.0049, 0.0008, 0.0312, ..., 0.0055, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0076, 0.0004, 0.0278], + ..., + [ 0.0170, -0.0082, 0.0471, ..., 0.0018, 0.0083, -0.0049], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0079, -0.0203], + [ 0.0033, 0.0081, -0.0077, ..., -0.0098, -0.0076, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[-0.0044, -0.0151, 0.0284, ..., -0.0109, 0.0140, -0.0002], + [ 0.0156, 0.0332, 0.0337, ..., 0.0223, -0.0165, -0.0015], + [-0.0071, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0068, 0.0062, 0.0280, ..., -0.0066, -0.0072, 0.0137], + [-0.0073, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0204], + [-0.0206, 0.0159, -0.0091, ..., 0.0045, 0.0167, -0.0086]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0055, ..., -0.0238, -0.0102, 0.0119], + [-0.0250, -0.0162, 0.0019, ..., -0.0084, 0.0035, -0.0369], + [ 0.0004, 0.0141, 0.0061, ..., -0.0452, -0.0056, 0.0121], + ..., + [ 0.0147, -0.0140, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0099, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0141, -0.0048, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[ 0.0013, -0.0327, -0.0190, ..., 0.0053, -0.0018, -0.0209], + [ 0.0713, 0.0150, -0.0026, ..., 0.0073, 0.0186, -0.0029], + [-0.0227, -0.0090, 0.0188, ..., -0.0125, 0.0193, -0.0049], + ..., + [ 0.0117, 0.0464, 0.0079, ..., 0.0248, 0.0168, -0.0038], + [ 0.0381, 0.0045, 0.0067, ..., 0.0048, -0.0110, 0.0035], + [ 0.0292, 0.0055, 0.0209, ..., 0.0044, 0.0234, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0063, 0.0212, ..., 0.0305, 0.0137, 0.0104], + [-0.0072, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0050], + ..., + [ 0.0082, -0.0275, 0.0225, ..., 0.0051, -0.0189, -0.0104], + [-0.0125, -0.0051, -0.0041, ..., 0.0046, -0.0048, 0.0211], + [ 0.0046, -0.0064, -0.0060, ..., -0.0109, 0.0264, 0.0047]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[-0.0002, -0.0076, -0.0109, ..., 0.0186, -0.0216, 0.0062], + [ 0.0045, 0.0123, -0.0231, ..., 0.0125, -0.0114, -0.0183], + [ 0.0107, 0.0103, -0.0038, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0121, 0.0146, ..., 0.0540, 0.0415, 0.0096], + [ 0.0061, 0.0084, -0.0143, ..., 0.0060, -0.0189, -0.0237], + [ 0.0048, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0067]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0019, 0.0359, ..., -0.0083, -0.0123, -0.0050], + [ 0.0190, -0.0044, 0.0298, ..., 0.0127, 0.0042, -0.0200], + [ 0.0114, -0.0112, 0.0177, ..., -0.0248, 0.0229, 0.0025], + ..., + [-0.0082, -0.0049, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0042, -0.0291, ..., 0.0206, 0.0304, -0.0089], + [-0.0081, 0.0245, -0.0066, ..., -0.0413, 0.0216, -0.0117]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0078, -0.0157, 0.0077, ..., -0.0049, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0064], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, 0.0038], + ..., + [-0.0036, 0.0003, 0.0258, ..., -0.0310, 0.0091, -0.0361], + [-0.0078, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [ 0.0067, 0.0157, -0.0070, ..., 0.0040, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0077, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0106, ..., 0.0103, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0041, ..., 0.0129, -0.0046, 0.0085], + ..., + [-0.0225, 0.0337, -0.0085, ..., -0.0177, 0.0542, -0.0249], + [ 0.0013, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0128, 0.0206, ..., 0.0071, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0095, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0052, 0.0063, ..., 0.0039, 0.0113, 0.0598], + [-0.0068, 0.0267, 0.0054, ..., 0.0284, -0.0060, -0.0364], + ..., + [ 0.0044, 0.0040, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0046, -0.0089, ..., 0.0211, 0.0135, -0.0075], + [ 0.0101, 0.0236, -0.0206, ..., -0.0135, -0.0048, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0067, 0.0294, ..., 0.0050, 0.0270, 0.0063], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0065, -0.0168], + [ 0.0201, -0.0120, 0.0126, ..., -0.0381, -0.0247, -0.0076], + ..., + [ 0.0275, -0.0182, -0.0183, ..., 0.0464, 0.0091, -0.0077], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0067, -0.0466], + [ 0.0327, 0.0284, -0.0030, ..., 0.0143, 0.0142, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, -0.0027, -0.0167, ..., -0.0039, 0.0170, 0.0107], + [-0.0002, -0.0002, 0.0295, ..., -0.0172, 0.0048, 0.0135], + [ 0.0193, 0.0236, -0.0123, ..., -0.0007, -0.0369, 0.0239], + ..., + [ 0.0150, -0.0413, -0.0126, ..., -0.0354, 0.0052, -0.0320], + [-0.0265, -0.0078, -0.0327, ..., -0.0070, 0.0077, 0.0074], + [ 0.0299, 0.0356, -0.0192, ..., 0.0059, -0.0052, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0013, 0.0041, 0.0317, ..., 0.0042, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0076, -0.0093], + [-0.0166, 0.0072, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0019, 0.0260, ..., 0.0159, 0.0048, -0.0474], + [-0.0056, 0.0080, 0.0077, ..., 0.0146, 0.0557, 0.0048], + [-0.0203, 0.0273, 0.0047, ..., -0.0330, 0.0065, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0058, ..., -0.0014, -0.0093, -0.0277], + [ 0.0060, -0.0078, -0.0062, ..., 0.0036, 0.0140, -0.0161], + [-0.0052, 0.0168, -0.0059, ..., -0.0092, 0.0124, 0.0043], + ..., + [ 0.0398, 0.0452, 0.0251, ..., 0.0039, -0.0330, 0.0051], + [-0.0620, -0.0039, 0.0371, ..., 0.0410, 0.0061, -0.0173], + [ 0.0127, 0.0074, -0.0270, ..., 0.0081, 0.0388, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0040, 0.0245], + [-0.0449, 0.0025, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0064, ..., 0.0046, 0.0288, 0.0125], + [ 0.0330, 0.0038, 0.0057, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0069, 0.0045, ..., -0.0077, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0070, 0.0038, 0.0527], + [-0.0051, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0095, -0.0096, 0.0327], + [ 0.0146, 0.0116, -0.0030, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0129, 0.0281, ..., -0.0242, -0.0126, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0155, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0041], + [-0.0048, -0.0045, -0.0286, ..., 0.0050, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0057, ..., -0.0019, 0.0082, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0096, 0.0150], + [-0.0071, 0.0029, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0063, ..., 0.0092, -0.0114, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0070, -0.0033, ..., -0.0220, -0.0145, -0.0046], + [-0.0166, 0.0189, -0.0058, ..., 0.0096, 0.0033, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0067, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0066, ..., 0.0024, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0069, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0121, 0.0289, 0.0126, ..., 0.0134, 0.0066, -0.0103], + [ 0.0188, -0.0197, -0.0022, ..., 0.0284, -0.0342, -0.0058], + [-0.0525, -0.0214, 0.0154, ..., -0.0270, 0.0081, 0.0254], + ..., + [-0.0052, 0.0087, 0.0281, ..., 0.0129, -0.0042, 0.0049], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0037, -0.0054], + [-0.0273, -0.0747, 0.0208, ..., -0.0042, -0.0052, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 9.1171e-04, 1.4954e-02, 3.3447e-02, ..., 2.0508e-02, + -2.7344e-02, -3.7537e-03], + [-1.6479e-02, 1.1108e-02, -1.3489e-02, ..., 1.0620e-02, + -7.2327e-03, 2.1851e-02], + [ 5.0659e-03, 2.2705e-02, -5.6458e-03, ..., 1.2817e-02, + -7.1106e-03, 1.4954e-02], + ..., + [-3.4668e-02, -1.4709e-02, 6.8359e-03, ..., 5.5847e-03, + -5.0659e-03, -9.2773e-03], + [-6.5613e-03, 3.9795e-02, -3.4424e-02, ..., 2.4536e-02, + 6.3477e-02, -1.1780e-02], + [-5.3024e-04, -5.0049e-02, -2.3193e-02, ..., 3.1982e-02, + -7.5340e-05, -5.5847e-03]], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0154, ..., -0.0344, 0.0067, -0.0364], + [ 0.0220, -0.0030, -0.0131, ..., 0.0078, 0.0101, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, 0.0072, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0073, 0.0192, -0.0393], + [-0.0105, -0.0244, 0.0275, ..., -0.0052, 0.0042, 0.0081], + [-0.0170, -0.0427, 0.0142, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0046, -0.0378, -0.0359, ..., 0.0552, -0.0078, -0.0160], + [ 0.0045, 0.0140, 0.0133, ..., 0.0050, -0.0153, 0.0354], + [ 0.0095, -0.0042, 0.0045, ..., -0.0047, -0.0249, -0.0136], + ..., + [ 0.0079, -0.0125, 0.0051, ..., -0.0095, -0.0084, 0.0052], + [-0.0060, 0.0439, 0.0166, ..., 0.0248, 0.0132, -0.0187], + [ 0.0175, 0.0063, 0.0044, ..., 0.0410, -0.0096, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0105, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0145, 0.0025, ..., 0.0065, -0.0244, 0.0356], + [ 0.0079, 0.0165, -0.0200, ..., -0.0080, 0.0237, -0.0081], + ..., + [-0.0197, -0.0022, 0.0386, ..., 0.0181, 0.0113, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, -0.0041, -0.0374], + [-0.0086, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0075, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0137, 0.0115, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, 0.0013, -0.0066, ..., 0.0050, -0.0045, 0.0337], + ..., + [ 0.0422, 0.0054, 0.0352, ..., -0.0304, -0.0035, -0.0204], + [ 0.0302, 0.0156, 0.0100, ..., -0.0049, -0.0254, -0.0245], + [-0.0067, -0.0064, 0.0078, ..., -0.0078, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0044, 0.0266, ..., 0.0154, 0.0187, 0.0049], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0096], + [-0.0189, 0.0205, 0.0044, ..., -0.0172, -0.0090, 0.0231], + ..., + [-0.0295, -0.0137, 0.0045, ..., 0.0039, -0.0050, 0.0099], + [-0.0052, -0.0140, 0.0317, ..., -0.0153, -0.0459, -0.0065], + [-0.0041, -0.0051, -0.0259, ..., -0.0040, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0059, -0.0140, ..., 0.0242, 0.0055, -0.0386], + [-0.0141, 0.0060, -0.0097, ..., -0.0369, -0.0155, 0.0157], + [-0.0267, -0.0041, -0.0065, ..., -0.0127, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0029, 0.0144, ..., 0.0115, 0.0040, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., 0.0045, -0.0038, -0.0137], + [-0.0219, -0.0396, -0.0010, ..., -0.0077, -0.0033, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0051, 0.0168, 0.0194, ..., 0.0349, -0.0420, -0.0048], + [ 0.0044, 0.0273, 0.0146, ..., -0.0184, 0.0011, 0.0156], + [ 0.0145, -0.0031, -0.0070, ..., 0.0245, 0.0359, 0.0061], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0037, 0.0043, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0072, -0.0312, -0.0457], + [ 0.0055, 0.0039, -0.0058, ..., -0.0038, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0047, 0.0121, 0.0038, ..., 0.0124, 0.0006, 0.0067], + [-0.0277, -0.0039, -0.0181, ..., 0.0005, 0.0020, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0038, 0.0513, -0.0113], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0113], + [-0.0282, -0.0040, 0.0121, ..., -0.0057, -0.0039, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, 0.0048, -0.0128, ..., 0.0140, 0.0308, -0.0066], + [ 0.0120, -0.0332, -0.0320, ..., -0.0045, -0.0140, -0.0039], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0088, 0.0295, 0.0505], + [-0.0159, 0.0025, -0.0311, ..., 0.0120, 0.0063, 0.0081], + [ 0.0044, 0.0198, 0.0141, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, 0.0043, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0090, -0.0040, 0.0052, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0045, ..., 0.0302, -0.0037, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0041, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0020, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0045, -0.0222, 0.0262], + ..., + [-0.0147, 0.0051, 0.0260, ..., 0.0045, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0057, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0042, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0112], + [-0.0278, -0.0054, -0.0374, ..., 0.0276, -0.0046, 0.0151], + [ 0.0089, 0.0237, -0.0076, ..., 0.0527, -0.0078, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0075, -0.0095], + [-0.0068, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0039, -0.0128, ..., 0.0060, -0.0371, -0.0046], + [ 0.0071, -0.0127, 0.0045, ..., -0.0204, 0.0039, 0.0049], + [-0.0110, -0.0195, -0.0050, ..., -0.0542, -0.0092, -0.0063], + ..., + [-0.0205, -0.0084, 0.0039, ..., -0.0199, 0.0046, 0.0039], + [-0.0220, -0.0302, -0.0101, ..., 0.0105, 0.0369, 0.0041], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[-0.0042, 0.0347, 0.0092, ..., 0.0277, -0.0208, -0.0220], + [-0.0038, 0.0043, 0.0238, ..., -0.0145, -0.0320, -0.0046], + [-0.0193, 0.0364, 0.0039, ..., -0.0077, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0065, -0.0232, -0.0464], + [-0.0110, -0.0298, -0.0195, ..., 0.0112, 0.0312, 0.0154], + [-0.0072, -0.0119, -0.0195, ..., 0.0041, 0.0109, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0078, -0.0239, -0.0123, ..., -0.0062, 0.0243, -0.0396], + [-0.0100, 0.0154, 0.0026, ..., -0.0277, -0.0027, -0.0038], + ..., + [ 0.0117, -0.0052, 0.0515, ..., 0.0327, -0.0120, -0.0074], + [ 0.0039, 0.0170, 0.0055, ..., 0.0039, 0.0280, -0.0210], + [-0.0025, 0.0072, 0.0045, ..., 0.0155, 0.0042, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [ 0.0040, 0.0081, -0.0074, ..., -0.0588, 0.0187, -0.0078], + [-0.0102, 0.0199, -0.0078, ..., 0.0080, -0.0332, -0.0076], + ..., + [-0.0052, -0.0464, 0.0277, ..., -0.0143, 0.0198, 0.0054], + [ 0.0162, 0.0011, 0.0085, ..., -0.0104, -0.0066, 0.0297], + [-0.0269, 0.0131, 0.0164, ..., 0.0084, 0.0176, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0046, ..., -0.0505, -0.0332, 0.0154], + [ 0.0208, 0.0366, 0.0007, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, 0.0058, -0.0201, ..., 0.0330, 0.0309, -0.0064], + ..., + [ 0.0028, 0.0123, 0.0153, ..., 0.0039, 0.0034, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0042, 0.0175], + [-0.0244, -0.0062, -0.0300, ..., -0.0061, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, 0.0040, 0.0075, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0050, ..., 0.0310, 0.0088, -0.0039], + [-0.0078, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0087, -0.0109, -0.0237, ..., -0.0104, 0.0157, -0.0021], + [-0.0085, 0.0205, 0.0145, ..., 0.0262, -0.0195, 0.0095], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0070]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0029, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0029, 0.0034, -0.0093], + [ 0.0017, -0.0011, 0.0398, ..., -0.0304, -0.0030, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0039], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0048, 0.0291], + [ 0.0132, -0.0004, -0.0339, ..., 0.0065, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, 0.0024], + [-0.0166, -0.0043, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0064, 0.0020, ..., 0.0240, 0.0031, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0020], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0009, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0066, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0041], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, 0.0034, 0.0021], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, -0.0042, ..., -0.0164, -0.0040, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0027, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0053, -0.0231, -0.0255, ..., -0.0068, 0.0295, -0.0046], + [-0.0255, 0.0513, 0.0030, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0035, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0064, -0.0066, -0.0364, ..., -0.0175, 0.0306, -0.0077], + [-0.0051, -0.0195, -0.0027, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0078, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0078, -0.0193, -0.0140], + [-0.0166, -0.0078, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0134, -0.0044, 0.0137, ..., 0.0021, 0.0211, -0.0275], + [ 0.0051, -0.0076, -0.0234, ..., 0.0088, -0.0084, -0.0166], + [ 0.0225, -0.0251, 0.0039, ..., 0.0072, 0.0236, 0.0077], + ..., + [ 0.0184, 0.0145, -0.0041, ..., 0.0269, -0.0085, -0.0096], + [-0.0038, 0.0044, 0.0078, ..., -0.0260, 0.0267, 0.0150], + [ 0.0102, 0.0043, 0.0212, ..., 0.0195, -0.0157, 0.0124]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0085, -0.0270, 0.0251, ..., 0.0095, 0.0052, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.0.proj.1.weight': tensor([[ 0.0018, 0.0251, 0.0957, ..., -0.0535, 0.0286, -0.0262], + [-0.0195, 0.0112, -0.1108, ..., 0.0304, -0.0312, 0.0593], + [ 0.0017, 0.1309, -0.0078, ..., -0.0058, 0.0078, -0.0282], + ..., + [ 0.0039, -0.0038, -0.0136, ..., 0.0142, -0.0222, -0.0051], + [-0.0349, -0.0645, 0.0508, ..., 0.0364, 0.0630, 0.0566], + [ 0.0192, -0.0223, 0.0010, ..., 0.0115, 0.0510, -0.0084]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_proj.weight': tensor([[ 0.0182, 0.0081, 0.0123, ..., 0.0016, -0.0015, 0.0272], + [ 0.0183, -0.0024, 0.0045, ..., 0.0009, -0.0028, 0.0209], + [-0.0061, 0.0053, -0.0056, ..., -0.0055, -0.0039, -0.0149], + ..., + [-0.0089, -0.0040, -0.0085, ..., 0.0106, -0.0037, -0.0076], + [-0.0016, -0.0033, -0.0052, ..., -0.0043, -0.0039, -0.0052], + [ 0.0081, 0.0026, 0.0023, ..., 0.0082, -0.0038, 0.0005]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_norm.weight': tensor([ 1.0742e-01, -9.8633e-02, 1.1279e-01, 3.6914e-01, -4.8056e-07, + -1.9092e-07, 2.0752e-03, 7.8870e-13, 5.4956e-15, 4.9738e-13, + -2.0961e-13, 3.3307e-16, 1.1519e-15, -4.4631e-14, 1.0868e-10, + -4.3485e-12, -3.3617e-05, 1.4827e-06, -8.6212e-04, 5.6744e-05, + 1.1505e-10, -2.5466e-11, 1.9409e-02, -4.1962e-04, 3.0518e-02, + 6.6406e-01, 1.3867e-01, -2.4414e-03, 5.6641e-02, 1.0596e-01, + 1.3279e-10, 5.4845e-14, -2.2736e-03, 1.4663e-05, 2.9755e-03, + 1.0376e-03, 6.8212e-11, -2.9650e-10, 4.8637e-05, 1.5099e-14, + 9.0639e-17, -1.2434e-14, 6.4393e-15, 4.4250e-03, 3.2471e-02, + 2.2125e-03, -3.5400e-02, -4.6875e-02, 4.8242e-01, 2.6703e-03, + -6.8665e-03, 3.4027e-03, 3.9062e-03, 1.6838e-06, 2.1210e-03, + 1.2733e-10, -1.3131e-11, 3.3396e-13, -4.5103e-16, 1.0249e-19, + -1.2967e-13, -3.8858e-15, 2.6557e-10, 3.3168e-15, 3.3203e-01, + 3.4766e-01, 4.3750e-01, 1.1768e-01, -2.4959e-07, -1.0245e-07, + 6.9580e-03, 1.5135e-12, -2.9088e-14, -3.2152e-13, 2.8866e-14, + 3.3307e-16, -1.2663e-16, -2.4425e-14, 1.8758e-11, -6.3949e-13, + 7.3910e-05, 4.0770e-05, 2.8133e-05, 5.7220e-05, -5.3478e-10, + -1.2296e-09, -8.8120e-04, -1.3580e-03, 4.8633e-01, 9.0820e-02, + 6.0938e-01, 7.9590e-02, 4.5471e-03, -2.6550e-03, 5.6389e-11, + 1.0081e-13, 2.9373e-04, 1.6117e-04, -9.7752e-06, 4.9210e-04, + -5.7116e-10, 1.4734e-10, 3.7994e-03, -3.6915e-15, 7.5287e-16, + -4.1980e-16, 2.0400e-15, 2.5269e-02, -6.6223e-03, -1.1978e-03, + 4.5117e-01, 2.8516e-01, 1.0498e-01, -7.8125e-03, 7.8125e-03, + -2.9907e-03, -4.0894e-03, 2.6375e-06, 2.1057e-03, -1.4552e-10, + 3.4333e-11, -4.5119e-13, 1.4502e-15, -3.6321e-18, 4.6896e-13, + -9.6367e-14, 4.2917e-12, 3.7303e-14], dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.self_attn.k_proj.weight': tensor([[-0.0102, 0.0103, -0.0096, ..., -0.0046, -0.0105, 0.0022], + [-0.0124, 0.0089, 0.0037, ..., 0.0029, -0.0098, 0.0126], + [-0.0073, -0.0295, 0.0043, ..., -0.0023, 0.0043, -0.0208], + ..., + [-0.0018, 0.0002, -0.0042, ..., -0.0025, -0.0005, -0.0027], + [-0.0027, -0.0030, -0.0047, ..., 0.0026, 0.0041, 0.0038], + [-0.0037, -0.0039, -0.0026, ..., 0.0029, -0.0006, 0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.k_norm.weight': tensor([ 2.9102e-01, 2.0312e-01, -2.0508e-01, 3.4961e-01, 4.7497e-07, + -2.1793e-07, 4.3335e-03, 8.1357e-13, 1.1990e-14, 5.0449e-13, + 1.8741e-13, 2.9837e-16, 1.1519e-15, -4.4853e-14, -1.0823e-10, + -4.4054e-12, 7.4863e-05, -1.2945e-07, 1.0910e-03, 5.6744e-05, + 1.2187e-10, -2.9786e-11, 1.3367e-02, 7.2861e-04, 4.1260e-02, + 6.6406e-01, 1.9336e-01, 7.6660e-02, 4.1809e-03, 2.9206e-05, + -1.3733e-10, 6.8390e-14, 2.2888e-03, 4.6492e-05, 2.9907e-03, + 1.0376e-03, 7.9126e-11, 2.9286e-10, 6.8665e-05, 1.4322e-14, + 1.1241e-15, -6.6058e-15, -6.4670e-15, 2.1118e-02, 2.9297e-02, + 2.1820e-03, 1.0254e-01, 1.1963e-01, 5.0391e-01, 7.8125e-03, + -7.8125e-03, -3.9062e-03, 4.0894e-03, -3.2783e-06, -2.0142e-03, + -6.5938e-11, -1.4353e-12, 3.4639e-13, 2.3419e-16, -4.7095e-19, + 1.4655e-13, -7.8271e-15, -2.6739e-10, 3.7470e-15, 1.7285e-01, + 3.4180e-01, 4.3164e-01, -2.6172e-01, -2.3562e-07, -1.1583e-08, + 3.5553e-03, 1.4566e-12, -2.6978e-14, 3.3751e-13, 3.1752e-14, + -3.8858e-16, 1.0278e-16, 1.5543e-14, -1.9554e-11, -6.6436e-13, + 1.1349e-04, -4.7684e-05, 1.4603e-05, 5.8413e-05, 5.3842e-10, + -1.2151e-09, 2.1744e-04, -9.0408e-04, 4.8242e-01, -8.9355e-02, + 6.0156e-01, -3.7689e-03, 5.5664e-02, 1.0547e-01, 1.1880e-11, + 2.2093e-14, -3.5095e-04, 1.6117e-04, -1.9169e-04, -4.8256e-04, + 5.7844e-10, -2.5466e-10, 3.7689e-03, 3.7748e-15, 5.4678e-15, + -8.6736e-17, 2.0817e-15, 1.4282e-02, 1.7334e-02, 1.4725e-03, + 4.5312e-01, 2.6172e-01, 1.4453e-01, 4.6082e-03, 7.6294e-03, + 1.3199e-03, 3.0975e-03, 9.3877e-07, 2.2583e-03, 1.9281e-10, + 3.6607e-11, 4.1389e-13, -1.4919e-15, 3.5508e-18, -4.6896e-13, + -9.6811e-14, 1.1084e-12, 3.7303e-14], dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.self_attn.v_proj.weight': tensor([[-0.0022, 0.0098, 0.0055, ..., -0.0032, -0.0040, -0.0063], + [ 0.0002, 0.0038, -0.0006, ..., -0.0014, 0.0046, 0.0002], + [ 0.0142, 0.0043, 0.0038, ..., -0.0041, 0.0016, 0.0063], + ..., + [-0.0006, 0.0021, -0.0019, ..., 0.0025, -0.0025, -0.0029], + [ 0.0063, -0.0032, 0.0037, ..., 0.0059, -0.0058, 0.0040], + [-0.0090, 0.0049, -0.0024, ..., 0.0055, -0.0038, -0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.output_proj.weight': tensor([[ 0.0145, -0.0216, -0.0175, ..., 0.0042, 0.0087, -0.0032], + [ 0.0022, -0.0068, 0.0124, ..., -0.0063, -0.0024, -0.0059], + [ 0.0040, 0.0022, 0.0052, ..., 0.0009, -0.0058, 0.0017], + ..., + [ 0.0055, -0.0001, 0.0048, ..., 0.0042, -0.0014, 0.0009], + [-0.0055, -0.0045, 0.0275, ..., 0.0036, -0.0033, 0.0161], + [-0.0084, 0.0095, -0.0052, ..., 0.0051, 0.0051, -0.0053]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_proj.weight': tensor([[-0.0034, -0.0093, -0.0017, ..., -0.0048, -0.0034, -0.0014], + [-0.0096, 0.0098, 0.0019, ..., 0.0072, 0.0025, -0.0030], + [-0.0023, -0.0084, 0.0025, ..., -0.0004, 0.0062, 0.0029], + ..., + [ 0.0030, 0.0071, 0.0009, ..., 0.0047, -0.0041, 0.0082], + [-0.0015, -0.0042, -0.0017, ..., -0.0048, 0.0043, -0.0067], + [-0.0060, 0.0023, 0.0009, ..., -0.0008, 0.0033, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_norm.weight': tensor([ 6.0120e-03, 8.8811e-06, 2.7740e-11, 1.4526e-02, -9.6436e-03, + 1.5442e-02, 2.2054e-06, 5.4550e-04, 3.4912e-02, 2.6978e-02, + 1.6880e-08, 2.8076e-02, 5.7812e-01, 9.3994e-03, 1.0437e-02, + 4.0894e-03, 4.6680e-01, 1.1086e-05, 1.1215e-03, 1.0803e-02, + 8.8501e-03, 7.3547e-03, 8.3496e-02, 1.8120e-05, 1.8433e-02, + 1.0681e-03, 7.9346e-03, -8.4400e-05, 1.2436e-03, 1.6708e-03, + 3.5156e-01, 9.0942e-03, 5.0306e-05, 4.6680e-01, 5.4297e-01, + 1.3828e-04, -5.1022e-05, 3.1494e-02, 1.5137e-01, 2.5635e-02, + 2.0996e-02, -4.6790e-06, 1.3642e-10, 4.1809e-03, 9.4223e-04, + 3.3203e-02, 1.2024e-02, 1.0605e-03, 4.0894e-03, 8.7891e-03, + 1.8539e-03, -7.4863e-05, -4.2915e-05, 1.5106e-03, -3.8300e-03, + 1.6689e-04, 1.9653e-02, 1.1396e-04, 1.7090e-02, 1.1475e-02, + 1.9932e-04, -7.8082e-06, 1.2329e-02, 7.2632e-03, 4.0293e-05, + 1.3550e-02, 9.9945e-04, 2.5000e-01, 8.8501e-04, 2.1851e-02, + 4.6875e-01, -1.2875e-05, 4.5410e-02, 8.4229e-03, 1.7695e-08, + 7.0190e-03, 1.6632e-03, 7.0801e-03, -1.3590e-05, 2.0752e-02, + 2.3438e-02, 4.2725e-03, 1.7188e-01, 7.6294e-03, 1.5869e-03, + 2.4567e-03, 1.3000e-02, -1.3530e-05, 1.0395e-04, 5.0781e-01, + 2.6550e-03, 8.9645e-05, 1.1520e-03, 3.8147e-06, 9.3384e-03, + 4.5117e-01, 4.3640e-03, 1.0538e-04, -3.7126e-13, 9.4223e-04, + 3.9062e-01, 1.3351e-05, -9.0332e-03, 1.7822e-02, 2.3071e-02, + 9.3460e-05, -1.5080e-05, 1.3769e-05, 1.6708e-03, 1.7944e-02, + 1.1597e-02, 8.5449e-04, 4.2419e-03, 1.2329e-02, -3.6316e-03, + 1.0395e-04, 1.1230e-02, -2.8534e-03, -1.1292e-03, 1.8311e-03, + 1.0300e-04, 6.3782e-03, 1.1368e-03, -2.2650e-05, 1.8978e-04, + 1.1414e-02, -7.1526e-06, 1.0133e-05], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_proj.weight': tensor([[ 1.5411e-03, 7.2861e-04, 1.9150e-03, ..., -6.7520e-04, + -2.4605e-04, 6.1035e-04], + [-3.6926e-03, -2.0599e-03, -3.9368e-03, ..., -2.6093e-03, + -3.9673e-03, -3.0060e-03], + [-1.7700e-03, -1.4877e-03, -2.4872e-03, ..., -2.8381e-03, + -5.6458e-03, -9.6130e-04], + ..., + [ 2.5635e-03, -2.4719e-03, 1.8768e-03, ..., 5.2643e-04, + 2.3041e-03, -5.7220e-04], + [-8.8882e-04, 5.3024e-04, -2.3193e-03, ..., -3.0212e-03, + 5.0354e-04, 2.3193e-03], + [ 2.9297e-03, 1.8997e-03, -3.2501e-03, ..., -4.2725e-03, + 6.5613e-04, 5.7697e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_norm.weight': tensor([ 5.9814e-03, -5.5730e-06, -2.7740e-11, 1.4343e-02, 9.7046e-03, + 1.5442e-02, 2.1756e-06, -5.4550e-04, 3.3936e-02, 2.7100e-02, + -5.9954e-09, 2.8320e-02, 5.8203e-01, 9.3384e-03, 9.8877e-03, + 4.0588e-03, 4.6680e-01, 1.1325e-05, 9.9182e-04, 1.0620e-02, + 8.7280e-03, 7.3853e-03, 8.3496e-02, 3.7670e-05, 1.8311e-02, + 1.0757e-03, 7.5073e-03, -8.3923e-05, 6.2561e-04, 1.6403e-03, + 3.5156e-01, 8.9111e-03, 5.0068e-05, 4.6680e-01, 5.4297e-01, + -2.0981e-04, 5.8413e-05, 3.1982e-02, 1.5137e-01, 2.5879e-02, + 2.0996e-02, 9.7275e-05, 1.9936e-09, 4.0283e-03, -2.2793e-04, + 3.2715e-02, 1.1902e-02, 1.0757e-03, 4.0283e-03, 8.7280e-03, + 1.8463e-03, -7.4863e-05, 4.2915e-05, 1.5030e-03, 3.9978e-03, + 2.1076e-04, 1.9531e-02, 1.3351e-04, 1.7334e-02, 1.1536e-02, + 1.9932e-04, -8.1062e-06, 1.2085e-02, 7.2632e-03, 4.0054e-05, + 1.3733e-02, 9.9182e-04, 2.5000e-01, 8.9645e-04, 2.1729e-02, + 4.6875e-01, 1.2696e-05, 4.3945e-02, 8.4229e-03, -1.9674e-08, + 7.1411e-03, 5.5313e-04, 7.0190e-03, -1.4782e-05, 2.0752e-02, + 2.3560e-02, 4.2419e-03, 1.7188e-01, 7.4463e-03, 1.6251e-03, + 2.4872e-03, 1.4038e-02, 1.3590e-05, 1.0490e-04, 5.0781e-01, + 2.6398e-03, -1.2398e-04, 7.5912e-04, 3.7253e-06, 9.3384e-03, + 4.5117e-01, -4.2419e-03, -9.1553e-05, 3.7126e-13, 9.1934e-04, + 3.9062e-01, 1.3351e-05, 1.0193e-02, 1.7822e-02, 2.3071e-02, + 9.2983e-05, 1.7524e-05, 1.3888e-05, 8.1253e-04, 1.7944e-02, + 1.1841e-02, 8.5449e-04, 4.2114e-03, 1.2268e-02, 3.6163e-03, + -1.6785e-04, 1.1292e-02, 2.8687e-03, 1.3123e-03, 1.8845e-03, + 9.7275e-05, 6.5918e-03, 1.1139e-03, -9.0599e-05, 1.0633e-04, + 1.1292e-02, -6.8247e-06, -3.7700e-06], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.v_proj.weight': tensor([[-9.8877e-03, 2.1820e-03, 1.2695e-02, ..., 2.8534e-03, + 2.6550e-03, -1.0193e-02], + [-2.2461e-02, 8.1635e-04, 7.7209e-03, ..., -3.2043e-03, + -6.1035e-03, 9.3937e-05], + [-1.2573e-02, 2.4567e-03, 2.9297e-03, ..., -2.2888e-03, + 1.3000e-02, -1.2268e-02], + ..., + [ 1.0925e-02, 1.8692e-03, 1.6861e-03, ..., 2.5940e-04, + 2.9755e-03, 1.8799e-02], + [-9.5215e-03, 2.5330e-03, 1.0315e-02, ..., 1.8463e-03, + -1.7700e-02, -3.7079e-03], + [-1.6479e-02, -2.8839e-03, 7.1716e-03, ..., -2.0599e-03, + -6.0425e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.output_proj.weight': tensor([[ 0.0064, -0.0018, 0.0056, ..., -0.0030, 0.0018, 0.0080], + [ 0.0080, 0.0066, -0.0012, ..., 0.0162, 0.0041, -0.0110], + [ 0.0052, -0.0030, -0.0008, ..., -0.0004, -0.0073, -0.0052], + ..., + [-0.0071, -0.0077, -0.0051, ..., -0.0079, -0.0047, -0.0085], + [ 0.0101, -0.0025, 0.0059, ..., -0.0142, -0.0183, 0.0055], + [ 0.0106, -0.0081, -0.0041, ..., -0.0070, -0.0037, -0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_img.weight': tensor([[ 0.0067, 0.0119, -0.0041, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0075, 0.0249, 0.0291], + [-0.0095, 0.0046, -0.0115, ..., -0.0342, -0.0327, 0.0061], + ..., + [ 0.0074, -0.0067, 0.0040, ..., 0.0082, 0.0286, -0.0165], + [ 0.0078, 0.0013, -0.0102, ..., -0.0100, 0.0142, -0.0309], + [ 0.0242, -0.0106, 0.0255, ..., -0.0227, -0.0043, -0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, -0.0010], + [ 0.0026, 0.0120, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0085, ..., -0.0057, -0.0046, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0039, 0.0054], + [-0.0160, 0.0055, -0.0036, ..., 0.0133, -0.0173, -0.0049], + [-0.0051, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img.weight': tensor([[-0.0510, -0.0021, -0.0364, ..., -0.0225, -0.0143, -0.0055], + [-0.0157, 0.0053, -0.0364, ..., 0.0099, 0.0117, -0.0036], + [ 0.0025, 0.0615, 0.0056, ..., -0.0205, -0.0306, -0.0114], + ..., + [ 0.0008, -0.0018, -0.0103, ..., -0.0211, -0.0012, 0.0288], + [ 0.0266, -0.0115, -0.0056, ..., -0.0420, -0.0282, -0.0008], + [ 0.0286, 0.0325, 0.0095, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.mlp.layer1.weight': tensor([[ 0.0082, 0.0159, 0.0067, ..., 0.0008, 0.0096, 0.0064], + [ 0.0050, -0.0152, -0.0035, ..., -0.0021, -0.0115, 0.0064], + [ 0.0062, -0.0057, -0.0039, ..., 0.0176, -0.0176, 0.0121], + ..., + [-0.0037, 0.0004, 0.0240, ..., 0.0054, 0.0075, 0.0029], + [-0.0030, 0.0186, 0.0068, ..., 0.0021, -0.0079, 0.0085], + [ 0.0050, 0.0027, -0.0013, ..., 0.0029, -0.0074, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.mlp.layer2.weight': tensor([[-0.0077, -0.0042, -0.0074, ..., -0.0038, -0.0261, -0.0078], + [ 0.0042, 0.0085, -0.0029, ..., -0.0162, -0.0216, 0.0073], + [-0.0062, 0.0092, 0.0090, ..., 0.0012, -0.0023, 0.0053], + ..., + [-0.0044, -0.0049, -0.0091, ..., -0.0069, -0.0026, -0.0082], + [ 0.0079, 0.0125, 0.0215, ..., 0.0008, 0.0056, 0.0114], + [ 0.0085, 0.0122, 0.0012, ..., 0.0050, -0.0021, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0610, -0.0593, -0.0579, ..., 0.0030, 0.0031, 0.0031], + [ 0.0591, 0.0583, 0.0562, ..., -0.0024, -0.0025, -0.0026], + [ 0.0583, 0.0576, 0.0564, ..., -0.0026, -0.0027, -0.0027], + ..., + [ 0.0630, 0.0615, 0.0598, ..., -0.0024, -0.0025, -0.0025], + [ 0.0598, 0.0583, 0.0566, ..., -0.0022, -0.0023, -0.0022], + [ 0.0598, 0.0583, 0.0571, ..., -0.0030, -0.0031, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0009, 0.0014, 0.0015, ..., 0.0050, 0.0036, 0.0003], + [-0.0025, -0.0015, -0.0016, ..., 0.0022, 0.0022, 0.0024], + [-0.0035, 0.0033, 0.0031, ..., 0.0059, 0.0060, 0.0034], + ..., + [-0.0432, 0.0500, 0.0493, ..., 0.0505, 0.0464, 0.0437], + [-0.0058, 0.0069, 0.0061, ..., 0.0053, 0.0044, 0.0057], + [ 0.0096, -0.0104, -0.0103, ..., -0.0099, -0.0096, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_cross_attn.1.weight': tensor([[-8.5449e-03, 4.4556e-03, 4.4861e-03, ..., -6.7520e-04, + -6.0654e-04, -7.7438e-04], + [-3.8818e-02, -3.7109e-02, -3.7109e-02, ..., 2.5177e-04, + 2.5940e-04, 2.6512e-04], + [ 4.1748e-02, 4.0283e-02, 4.0039e-02, ..., -7.0095e-05, + -5.1022e-05, -7.5340e-05], + ..., + [ 3.9062e-02, 3.8086e-02, 3.7109e-02, ..., -3.3569e-04, + -2.6703e-04, -3.0327e-04], + [ 4.9072e-02, 4.8096e-02, 4.6631e-02, ..., 5.0354e-04, + 6.1417e-04, 5.4550e-04], + [ 3.8818e-02, 3.7598e-02, 3.6621e-02, ..., -5.6624e-06, + 7.0572e-05, 2.1935e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0129, -0.0072, 0.0067, ..., 0.0072, 0.0072, 0.0068], + [ 0.0052, 0.0034, -0.0049, ..., -0.0034, -0.0034, -0.0048], + [-0.0037, -0.0036, 0.0035, ..., 0.0036, 0.0035, 0.0036], + ..., + [-0.0063, -0.0205, 0.0364, ..., 0.0203, 0.0432, 0.0320], + [ 0.0103, 0.0139, 0.0101, ..., -0.0131, 0.0050, 0.0099], + [-0.0137, -0.0046, -0.0113, ..., 0.0045, -0.0094, -0.0110]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_mlp.1.weight': tensor([[-0.0049, -0.0049, 0.0038, ..., 0.0004, 0.0007, 0.0006], + [ 0.0640, 0.0618, 0.0605, ..., -0.0024, -0.0027, -0.0027], + [ 0.0554, 0.0542, 0.0525, ..., -0.0037, -0.0040, -0.0040], + ..., + [ 0.0952, 0.0928, 0.0908, ..., -0.0008, -0.0009, -0.0010], + [-0.0038, -0.0038, -0.0038, ..., -0.0004, -0.0005, -0.0004], + [ 0.0747, 0.0723, 0.0708, ..., -0.0012, -0.0012, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_mlp.2.weight': tensor([[-0.0077, 0.0091, 0.0107, ..., 0.0072, 0.0082, 0.0060], + [-0.0007, 0.0023, 0.0033, ..., 0.0031, 0.0007, 0.0009], + [-0.0017, 0.0060, 0.0061, ..., 0.0044, 0.0015, 0.0057], + ..., + [ 0.0036, 0.0747, 0.1025, ..., 0.0635, -0.0044, 0.0496], + [ 0.0114, 0.0165, 0.0096, ..., 0.0280, -0.0114, 0.0233], + [-0.0089, 0.0033, -0.0003, ..., 0.0041, 0.0087, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.before_proj.weight': tensor([[ 0.0028, 0.0002, -0.0006, ..., 0.0062, 0.0011, -0.0013], + [-0.0094, -0.0055, -0.0041, ..., -0.0078, -0.0024, 0.0052], + [-0.0075, -0.0059, -0.0033, ..., -0.0015, -0.0036, -0.0027], + ..., + [-0.0011, 0.0017, -0.0047, ..., 0.0033, -0.0049, -0.0004], + [-0.0037, -0.0017, 0.0011, ..., -0.0020, -0.0052, -0.0042], + [ 0.0027, 0.0028, -0.0021, ..., 0.0045, 0.0005, -0.0034]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.before_proj.bias': tensor([-0.0199, 0.0027, 0.0018, ..., 0.0016, -0.0040, -0.0239], + dtype=torch.bfloat16), 'net.control_blocks_0.0.after_proj.weight': tensor([[-0.0188, -0.0083, 0.0017, ..., -0.0049, -0.0119, 0.0113], + [ 0.0189, -0.0134, 0.0063, ..., -0.0007, 0.0084, -0.0087], + [ 0.0009, -0.0009, -0.0043, ..., -0.0043, 0.0015, -0.0023], + ..., + [ 0.0029, -0.0039, 0.0083, ..., -0.0016, -0.0035, 0.0015], + [ 0.0276, -0.0150, 0.0077, ..., -0.0062, -0.0171, -0.0134], + [-0.0036, -0.0123, 0.0019, ..., 0.0026, -0.0048, 0.0112]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.after_proj.bias': tensor([ 0.0075, -0.0165, -0.0026, ..., -0.0014, -0.0140, 0.0598], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_proj.weight': tensor([[-0.0220, 0.0138, -0.0125, ..., -0.0093, -0.0069, 0.0132], + [ 0.0049, 0.0032, -0.0035, ..., 0.0193, -0.0050, -0.0388], + [ 0.0012, 0.0103, 0.0013, ..., -0.0023, 0.0043, -0.0078], + ..., + [ 0.0009, 0.0050, -0.0006, ..., 0.0024, -0.0010, -0.0002], + [ 0.0021, -0.0055, -0.0013, ..., 0.0013, -0.0062, 0.0017], + [ 0.0025, -0.0006, 0.0052, ..., 0.0031, -0.0019, -0.0037]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_norm.weight': tensor([ 3.9062e-01, 1.2578e+00, 5.6396e-02, 1.1953e+00, 8.9844e-01, + 4.6945e-34, -1.9184e-34, -1.5573e-34, -6.8310e-34, 3.3087e-22, + 8.9676e-34, -4.9653e-34, -9.3287e-35, -3.7014e-34, -7.2523e-34, + -1.0653e-33, 1.4551e-01, 3.8338e-04, -2.1696e-05, 2.2339e-02, + 2.4319e-04, -5.7697e-05, 4.6289e-01, 8.8672e-01, 1.3516e+00, + 4.4531e-01, -8.6426e-02, 6.0547e-01, 8.4229e-03, -2.5022e-30, + 1.9259e-34, 1.6852e-33, -2.3127e-05, 8.9264e-04, 8.7738e-05, + -9.7656e-03, 3.1281e-04, -1.0909e-34, -4.9277e-35, 1.1977e-33, + -1.3542e-33, 2.1507e-22, 6.3496e-34, 1.0859e+00, 3.2422e-01, + 2.1973e-01, 3.7500e-01, 1.3438e+00, 2.0630e-02, -4.4556e-03, + -6.2256e-03, 2.3926e-01, 1.7395e-03, -4.6042e-34, -3.7917e-34, + -3.0333e-33, 7.4158e-03, -6.0100e-25, 1.9019e-33, 2.0613e-34, + 3.8556e-35, 4.0813e-35, -1.6325e-34, -2.7836e-35, 5.0391e-01, + 4.1406e-01, 6.3281e-01, 4.9805e-01, 1.6797e-01, 5.2662e-34, + -1.2883e-35, 4.4838e-34, 7.4028e-34, -3.2467e-23, 2.1667e-34, + -1.5528e-33, 2.7836e-35, -1.0833e-34, 9.3287e-34, 8.1852e-34, + 4.6387e-03, 8.9264e-04, -4.4346e-05, 4.4250e-04, 2.3079e-04, + 3.5889e-02, 1.1953e+00, 2.5391e-01, 4.5117e-01, 1.3359e+00, + 2.7734e-01, 2.4414e-01, 1.9455e-03, -2.0338e-30, 2.1215e-34, + -1.3120e-33, -3.0279e-05, 3.4714e-04, -1.6212e-05, 1.3550e-02, + 3.8719e-04, -1.1887e-34, -4.1001e-35, -1.6777e-34, -1.1857e-33, + -4.1359e-24, 4.2130e-33, 2.8711e-01, 9.4141e-01, 1.2266e+00, + 1.4844e+00, 1.9336e-01, 8.3594e-01, -3.9062e-03, -8.2397e-04, + -4.7913e-03, 2.5558e-04, 1.3181e-33, -3.4005e-34, 2.1787e-33, + 6.7520e-04, -7.9616e-24, 2.3232e-33, 1.7379e-34, -1.0909e-35, + 1.1134e-34, -6.9213e-35, 6.3947e-35], dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.self_attn.k_proj.weight': tensor([[ 0.0016, -0.0195, 0.0078, ..., 0.0172, -0.0051, -0.0264], + [ 0.0066, -0.0129, -0.0035, ..., -0.0199, 0.0034, -0.0095], + [ 0.0092, 0.0063, -0.0018, ..., 0.0054, -0.0044, 0.0065], + ..., + [-0.0085, 0.0008, -0.0060, ..., 0.0082, 0.0109, -0.0054], + [-0.0095, -0.0052, 0.0011, ..., -0.0036, 0.0062, -0.0051], + [-0.0058, -0.0028, -0.0025, ..., -0.0032, 0.0117, -0.0054]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.k_norm.weight': tensor([ 5.0391e-01, 1.2812e+00, 2.3438e-01, 1.2109e+00, 8.9844e-01, + 4.5139e-34, 1.3918e-34, 1.1586e-34, 1.5167e-33, 3.3087e-22, + -2.8137e-34, -2.0162e-34, 4.4387e-35, 1.9410e-34, -6.7708e-34, + 6.0938e-35, 1.4551e-01, 3.6812e-04, 1.2875e-04, 5.5420e-02, + 2.4319e-04, 1.2493e-04, 6.8750e-01, 8.9062e-01, 1.3516e+00, + 4.5117e-01, 2.1289e-01, 6.4453e-01, -8.8501e-03, 1.6147e-30, + 1.0382e-34, -7.8993e-35, 1.4782e-04, 8.6975e-04, 1.5545e-04, + 4.5410e-02, 3.0708e-04, -5.0781e-35, -2.4262e-35, 4.1829e-34, + -3.9120e-34, -8.5200e-23, 4.7245e-34, 1.0781e+00, 3.8672e-01, + -6.3477e-02, -3.8672e-01, 1.3359e+00, 1.5527e-01, 2.4567e-03, + -6.1646e-03, 2.3340e-01, 1.7395e-03, -2.5278e-34, 1.2940e-34, + 1.1857e-33, 7.1411e-03, -6.0746e-25, -3.3102e-34, 1.3015e-34, + -1.2413e-35, -6.1314e-35, 1.6852e-34, -2.2569e-35, 5.1562e-01, + 5.5078e-01, 6.2891e-01, 4.3359e-01, -2.4414e-01, -6.9514e-34, + 1.1360e-34, 2.0162e-34, -3.6352e-33, 1.5406e-23, -2.9190e-34, + -4.9051e-34, 1.9184e-35, 8.0874e-35, -2.1366e-34, 5.5972e-34, + 5.1575e-03, 8.8501e-04, 1.3161e-04, 4.2725e-04, 2.3460e-04, + 3.5889e-02, 1.1875e+00, 3.1836e-01, 4.7070e-01, 1.3438e+00, + 2.7539e-01, 3.0273e-01, 8.3923e-04, -1.8612e-30, 1.5799e-34, + -3.1296e-33, 1.0633e-04, 4.0627e-04, 4.3869e-05, 4.4678e-02, + 3.8528e-04, 3.6864e-35, -4.5515e-35, -1.3993e-34, -1.0472e-33, + 3.0632e-24, 2.0222e-33, 4.6094e-01, 9.4531e-01, 1.2344e+00, + 1.4844e+00, 2.3438e-01, 8.4375e-01, -6.0425e-03, -7.6294e-05, + -2.0703e-01, 3.1090e-04, -6.5000e-34, -3.7014e-34, -1.4264e-33, + 1.1587e-04, 3.5931e-24, 4.6945e-34, -1.0006e-34, -3.7616e-35, + -1.0382e-34, 3.2162e-35, -4.7396e-35], dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.self_attn.v_proj.weight': tensor([[ 0.0028, -0.0034, 0.0063, ..., -0.0203, -0.0049, 0.0070], + [ 0.0029, -0.0100, 0.0068, ..., 0.0069, -0.0041, -0.0063], + [ 0.0052, -0.0027, -0.0053, ..., -0.0231, -0.0046, -0.0111], + ..., + [ 0.0065, 0.0164, 0.0194, ..., -0.0010, 0.0004, -0.0175], + [-0.0047, -0.0121, 0.0077, ..., 0.0072, -0.0036, 0.0110], + [ 0.0112, -0.0040, -0.0160, ..., 0.0114, 0.0090, 0.0106]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.output_proj.weight': tensor([[-0.0054, -0.0135, -0.0113, ..., 0.0189, -0.0115, 0.0073], + [ 0.0178, -0.0056, -0.0044, ..., -0.0254, -0.0120, -0.0076], + [-0.0027, 0.0036, 0.0038, ..., -0.0048, 0.0053, 0.0042], + ..., + [ 0.0027, -0.0137, -0.0098, ..., -0.0210, -0.0160, -0.0188], + [ 0.0042, -0.0032, -0.0053, ..., -0.0211, 0.0063, -0.0063], + [ 0.0143, 0.0056, -0.0114, ..., -0.0033, 0.0105, -0.0081]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_proj.weight': tensor([[-0.0164, 0.0239, 0.0057, ..., 0.0036, -0.0100, 0.0140], + [ 0.0244, -0.0101, -0.0052, ..., -0.0085, 0.0110, -0.0208], + [ 0.0104, 0.0075, -0.0023, ..., -0.0057, 0.0073, -0.0005], + ..., + [ 0.0092, -0.0036, 0.0033, ..., 0.0065, -0.0099, 0.0070], + [-0.0061, -0.0043, -0.0024, ..., -0.0044, -0.0095, 0.0100], + [ 0.0219, 0.0141, -0.0095, ..., -0.0045, 0.0086, -0.0172]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_norm.weight': tensor([ 2.3682e-02, 4.8047e-01, 1.7822e-02, 8.7891e-03, 6.7749e-03, + 2.0606e-12, 9.5367e-06, 5.5469e-01, -1.8332e-12, 4.4727e-01, + 1.8799e-02, 1.9287e-02, 7.0496e-03, 2.2095e-02, 1.7944e-02, + 1.7456e-02, 6.0425e-03, 1.8215e-04, -3.6001e-05, 2.9297e-02, + 2.6562e-01, 2.2461e-02, 6.5231e-04, 1.8433e-02, 1.0757e-03, + 2.6131e-04, -1.7548e-04, 3.0884e-02, 6.5327e-05, 4.2969e-01, + -1.1969e-04, -4.2915e-05, 9.6436e-03, 1.4114e-04, 7.6904e-03, + 1.0376e-02, 2.0146e-05, 4.5703e-01, 1.5259e-02, 2.7771e-03, + 3.6377e-02, 1.8921e-02, 5.9204e-03, 2.8133e-05, 5.4321e-03, + 7.6294e-03, 1.8616e-03, 1.7578e-02, 2.8076e-02, 1.4893e-02, + 4.9438e-03, 4.4336e-01, 1.0071e-02, 1.5869e-02, 3.1494e-02, + 1.8692e-04, 6.8359e-02, 1.6479e-02, 1.5869e-02, 2.5391e-02, + 4.0770e-05, 1.1475e-02, 5.7373e-03, -4.8876e-05, 1.2756e-02, + 2.5482e-03, -1.8597e-05, -2.7776e-05, 8.9722e-03, 4.7852e-02, + -5.2387e-08, 6.1951e-03, 4.6143e-02, 3.2471e-02, 4.3640e-03, + 8.3008e-03, 2.9755e-03, 1.7578e-02, 2.1973e-02, 1.4648e-02, + -2.2793e-04, 1.1963e-02, 3.2471e-02, -2.8253e-05, 2.9419e-02, + 1.2360e-03, 4.3030e-03, 3.0518e-02, 1.1658e-02, 5.5847e-03, + 1.3306e-02, 5.9891e-04, 1.1292e-02, 2.1729e-02, 1.3306e-02, + -1.0395e-04, 1.9287e-02, 3.3203e-02, 1.5991e-02, 4.1748e-02, + 2.5513e-02, 2.3438e-02, 1.6357e-02, 1.0443e-04, -2.0373e-10, + 1.3733e-02, 4.6158e-04, 2.2461e-02, 1.6479e-02, 1.9531e-02, + 1.3184e-02, 5.8594e-03, 2.1118e-02, 5.0781e-01, 1.1063e-03, + 1.1353e-02, 7.8735e-03, 1.2512e-02, 5.1575e-03, 5.2002e-02, + 2.3499e-03, 2.3651e-03, -1.4210e-04, 1.2451e-02, 6.4087e-03, + 1.3367e-02, 2.0264e-02, 2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_proj.weight': tensor([[-0.0010, -0.0033, 0.0038, ..., -0.0006, 0.0034, 0.0026], + [ 0.0116, -0.0035, -0.0040, ..., 0.0050, 0.0197, 0.0012], + [ 0.0023, 0.0036, 0.0016, ..., -0.0006, -0.0019, 0.0029], + ..., + [-0.0017, -0.0049, -0.0036, ..., -0.0004, -0.0015, 0.0031], + [-0.0016, -0.0053, -0.0125, ..., 0.0029, 0.0045, 0.0010], + [ 0.0078, -0.0004, -0.0020, ..., -0.0027, 0.0015, -0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_norm.weight': tensor([ 2.4658e-02, 4.8047e-01, 1.7578e-02, 8.8501e-03, 6.7749e-03, + 2.1885e-12, 9.4771e-06, 5.5469e-01, 1.8474e-12, 4.4727e-01, + 1.9409e-02, 1.9165e-02, 7.0190e-03, 2.1973e-02, 1.7822e-02, + 1.7334e-02, 6.2866e-03, 1.7929e-04, -3.7432e-05, 2.9175e-02, + 2.6562e-01, 2.1973e-02, 6.5613e-04, 1.8555e-02, 1.0986e-03, + 2.6321e-04, -1.7357e-04, 3.0640e-02, 6.5327e-05, 4.2969e-01, + -1.1921e-04, -4.4823e-05, 9.4604e-03, 1.4019e-04, 7.6904e-03, + 1.0315e-02, 2.0504e-05, 4.5703e-01, 1.5259e-02, 2.6550e-03, + 3.5400e-02, 1.8921e-02, 5.5847e-03, 2.5511e-05, 5.3406e-03, + 7.6904e-03, 1.8768e-03, 1.7334e-02, 2.8198e-02, 1.5015e-02, + 4.8828e-03, 4.4336e-01, 1.0132e-02, 1.5747e-02, 3.1250e-02, + 1.8597e-04, 6.8848e-02, 1.6235e-02, 1.5869e-02, 2.5513e-02, + 4.1485e-05, 1.1536e-02, 5.7373e-03, -5.0783e-05, 1.2756e-02, + 2.4872e-03, -1.9431e-05, -2.7418e-05, 9.0942e-03, 4.7852e-02, + -1.1316e-07, 6.3477e-03, 4.6631e-02, 3.2471e-02, 4.4250e-03, + 8.6670e-03, 2.9297e-03, 1.7090e-02, 2.1973e-02, 1.4771e-02, + -2.2316e-04, 1.1780e-02, 3.1982e-02, -2.7776e-05, 2.9663e-02, + 1.0910e-03, 4.2725e-03, 3.0273e-02, 1.1963e-02, 5.7373e-03, + 1.3794e-02, 6.2561e-04, 1.1292e-02, 2.1606e-02, 1.3123e-02, + -1.0586e-04, 1.8677e-02, 3.2959e-02, 1.5869e-02, 4.1016e-02, + 2.6367e-02, 2.3560e-02, 1.6602e-02, 1.0347e-04, 2.0373e-10, + 1.3672e-02, 4.5013e-04, 2.2217e-02, 1.6602e-02, 1.9287e-02, + 1.3184e-02, 5.5847e-03, 2.0996e-02, 5.0781e-01, 1.0986e-03, + 1.1292e-02, 7.8735e-03, 1.2329e-02, 5.3711e-03, 5.2002e-02, + 2.3956e-03, 2.3956e-03, -1.4210e-04, 1.2634e-02, 6.3477e-03, + 1.3245e-02, 2.0264e-02, 2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.v_proj.weight': tensor([[-2.2583e-02, 2.2736e-03, 2.2430e-03, ..., -8.6784e-05, + -1.5442e-02, -2.7832e-02], + [-4.1199e-03, 1.0803e-02, 1.0376e-02, ..., 7.7820e-03, + 6.6833e-03, 7.7209e-03], + [-3.9307e-02, -5.0049e-03, -7.4158e-03, ..., -9.3994e-03, + -2.7710e-02, -1.2451e-02], + ..., + [-4.1748e-02, 7.0190e-04, -1.5625e-02, ..., 1.8158e-03, + 2.1729e-02, 9.3994e-03], + [ 6.5308e-03, 6.1646e-03, -4.4556e-03, ..., 3.6469e-03, + 8.6670e-03, 8.1787e-03], + [ 1.6235e-02, 4.6692e-03, -3.8757e-03, ..., 2.0294e-03, + 1.1353e-02, -4.6015e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.output_proj.weight': tensor([[ 0.0071, 0.0040, 0.0148, ..., 0.0079, 0.0043, -0.0017], + [-0.0080, -0.0045, -0.0041, ..., -0.0087, 0.0038, 0.0112], + [-0.0005, 0.0038, -0.0038, ..., 0.0015, -0.0027, -0.0022], + ..., + [ 0.0054, 0.0052, 0.0035, ..., -0.0064, -0.0051, 0.0078], + [ 0.0128, 0.0302, -0.0167, ..., -0.0222, 0.0045, 0.0425], + [-0.0012, -0.0087, -0.0229, ..., -0.0104, 0.0020, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_img.weight': tensor([[ 0.0079, 0.0020, 0.0430, ..., -0.0327, 0.0054, 0.0148], + [-0.0278, 0.0280, 0.0249, ..., -0.0068, -0.0378, 0.0271], + [ 0.0139, -0.0116, 0.0312, ..., -0.0075, 0.0253, -0.0110], + ..., + [ 0.0245, -0.0150, 0.0033, ..., -0.0452, 0.0216, -0.0209], + [-0.0081, 0.0073, -0.0471, ..., -0.0109, 0.0157, -0.0088], + [ 0.0078, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0038, -0.0123, -0.0115], + [ 0.0135, 0.0061, -0.0188, ..., 0.0157, 0.0082, -0.0042], + [-0.0306, -0.0028, -0.0272, ..., 0.0101, 0.0337, 0.0039], + ..., + [-0.0040, -0.0040, 0.0047, ..., 0.0209, -0.0103, 0.0073], + [ 0.0193, 0.0210, 0.0123, ..., -0.0145, 0.0117, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0110, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img.weight': tensor([[ 0.0058, -0.0056, 0.0173, ..., 0.0083, 0.0356, 0.0217], + [-0.0498, -0.0051, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0108, -0.0044, -0.0139, ..., 0.0039, 0.0269, 0.0077], + ..., + [-0.0043, 0.0125, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0077, -0.0162, ..., -0.0054, -0.0075, -0.0121], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0138, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.mlp.layer1.weight': tensor([[-0.0176, 0.0156, -0.0038, ..., -0.0092, -0.0087, -0.0221], + [-0.0026, -0.0106, 0.0096, ..., -0.0039, -0.0060, -0.0166], + [-0.0065, 0.0032, 0.0022, ..., -0.0078, -0.0032, -0.0050], + ..., + [ 0.0251, 0.0005, -0.0049, ..., 0.0010, 0.0085, 0.0071], + [-0.0251, 0.0047, -0.0145, ..., 0.0160, -0.0053, 0.0062], + [ 0.0092, -0.0067, -0.0008, ..., -0.0135, 0.0120, -0.0248]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.mlp.layer2.weight': tensor([[-0.0166, -0.0135, 0.0115, ..., -0.0025, 0.0060, -0.0061], + [-0.0049, 0.0071, 0.0084, ..., -0.0048, 0.0110, 0.0041], + [-0.0077, -0.0070, 0.0029, ..., -0.0098, 0.0082, -0.0077], + ..., + [-0.0027, 0.0090, 0.0045, ..., 0.0143, -0.0068, 0.0093], + [ 0.0049, 0.0103, 0.0046, ..., -0.0059, -0.0005, 0.0159], + [ 0.0105, -0.0038, -0.0099, ..., 0.0050, -0.0029, -0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0447, -0.0435, -0.0430, ..., 0.0050, 0.0053, 0.0053], + [ 0.0408, 0.0396, 0.0381, ..., -0.0037, -0.0036, -0.0037], + [-0.0454, -0.0437, -0.0432, ..., 0.0051, 0.0052, 0.0051], + ..., + [-0.0498, -0.0488, -0.0476, ..., 0.0035, 0.0036, 0.0037], + [-0.0439, -0.0432, -0.0425, ..., 0.0037, 0.0040, 0.0040], + [-0.0310, -0.0295, -0.0284, ..., 0.0011, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0058, 0.0006, 0.0046, ..., 0.0025, 0.0046, -0.0006], + [ 0.0018, -0.0023, -0.0004, ..., 0.0001, 0.0035, 0.0018], + [-0.0065, -0.0017, -0.0055, ..., -0.0077, -0.0070, 0.0017], + ..., + [-0.0293, 0.0015, -0.0344, ..., -0.0378, -0.0361, -0.0017], + [-0.0250, -0.0064, -0.0226, ..., -0.0249, -0.0222, 0.0072], + [ 0.0065, -0.0074, 0.0048, ..., 0.0065, -0.0005, 0.0062]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_cross_attn.1.weight': tensor([[ 6.6406e-02, 6.4941e-02, 6.3965e-02, ..., 4.8161e-05, + 1.6093e-05, 5.9843e-05], + [ 4.4556e-03, 4.4250e-03, 4.4250e-03, ..., -3.4332e-05, + -1.1063e-04, -2.2602e-04], + [-2.2827e-02, -2.1973e-02, -2.0996e-02, ..., -6.8283e-04, + -6.8283e-04, -6.9427e-04], + ..., + [-2.7344e-02, -2.6978e-02, -2.6245e-02, ..., -4.1580e-04, + -3.7575e-04, -3.3760e-04], + [-2.5391e-02, -2.4902e-02, -2.4170e-02, ..., -2.0599e-04, + -1.6022e-04, -2.1267e-04], + [ 5.8838e-02, 5.7617e-02, 5.6885e-02, ..., -1.4591e-04, + -2.6822e-05, -1.7166e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0094, -0.0076, -0.0112, ..., -0.0010, 0.0064, -0.0079], + [ 0.0042, 0.0039, 0.0047, ..., -0.0049, -0.0038, 0.0056], + [ 0.0130, -0.0038, -0.0038, ..., 0.0047, -0.0058, 0.0097], + ..., + [ 0.0679, 0.0107, -0.0051, ..., 0.0074, -0.0149, 0.0483], + [ 0.0288, -0.0056, -0.0077, ..., -0.0071, -0.0209, 0.0310], + [-0.0078, 0.0099, -0.0190, ..., 0.0144, 0.0126, -0.0093]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_mlp.1.weight': tensor([[-0.0043, -0.0045, -0.0044, ..., 0.0003, 0.0003, 0.0003], + [ 0.0659, 0.0640, 0.0635, ..., -0.0038, -0.0039, -0.0038], + [ 0.0415, 0.0408, 0.0403, ..., -0.0039, -0.0039, -0.0039], + ..., + [ 0.0320, 0.0309, 0.0306, ..., -0.0021, -0.0022, -0.0021], + [-0.0124, -0.0114, -0.0105, ..., 0.0014, 0.0014, 0.0013], + [ 0.0254, 0.0253, 0.0253, ..., -0.0025, -0.0026, -0.0026]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_mlp.2.weight': tensor([[-0.0028, 0.0018, 0.0028, ..., 0.0026, 0.0021, 0.0024], + [-0.0002, 0.0020, 0.0021, ..., 0.0019, -0.0001, 0.0010], + [ 0.0019, 0.0001, -0.0037, ..., -0.0031, -0.0019, -0.0029], + ..., + [-0.0029, 0.0013, -0.0189, ..., -0.0081, 0.0028, -0.0057], + [ 0.0151, -0.0075, -0.0244, ..., -0.0165, -0.0144, -0.0148], + [ 0.0172, -0.0195, -0.0133, ..., -0.0166, -0.0178, -0.0162]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.after_proj.weight': tensor([[ 0.0104, -0.0075, 0.0262, ..., -0.0078, 0.0193, -0.0015], + [-0.0023, -0.0077, 0.0088, ..., 0.0153, 0.0086, -0.0063], + [ 0.0081, 0.0028, -0.0168, ..., 0.0070, 0.0148, -0.0087], + ..., + [-0.0050, 0.0074, 0.0249, ..., -0.0206, -0.0128, -0.0078], + [-0.0036, 0.0177, 0.0115, ..., -0.0117, 0.0003, 0.0161], + [-0.0123, -0.0178, -0.0145, ..., 0.0075, 0.0110, -0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.after_proj.bias': tensor([-4.2419e-03, 5.1880e-03, 4.8523e-03, ..., 6.7749e-03, + 7.0572e-05, -1.3885e-03], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_proj.weight': tensor([[ 0.0249, -0.0265, -0.0237, ..., 0.0074, -0.0041, 0.0330], + [ 0.0048, 0.0044, -0.0016, ..., -0.0142, 0.0068, -0.0049], + [ 0.0253, 0.0031, -0.0038, ..., 0.0045, -0.0076, 0.0183], + ..., + [-0.0087, -0.0070, 0.0021, ..., -0.0044, -0.0036, -0.0004], + [ 0.0315, 0.0259, -0.0068, ..., -0.0079, -0.0048, 0.0042], + [-0.0124, -0.0121, 0.0179, ..., -0.0028, 0.0063, -0.0244]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_norm.weight': tensor([ 1.3750e+00, 5.2344e-01, 6.6406e-01, 4.7852e-01, 1.3359e+00, + 1.4893e-02, 9.6436e-03, 1.9789e-05, 2.1057e-03, 1.5137e-02, + 3.3677e-06, -7.6252e-09, 1.1406e+00, -2.5988e-05, -3.5508e-18, + -4.7119e-02, 1.1597e-02, 2.6489e-02, 2.6703e-04, 1.0010e-02, + 1.5527e-01, -3.7909e-05, 9.6875e-01, 1.5938e+00, 7.1094e-01, + 1.1016e+00, 7.1094e-01, 1.9141e-01, 2.3682e-02, -1.4099e-02, + 7.1106e-03, 3.1090e-04, 7.5989e-03, -3.0212e-03, -1.0071e-03, + 2.9663e-02, 1.3965e-01, -2.5272e-05, 5.9814e-03, 2.1973e-03, + -7.0930e-06, -6.2883e-06, -1.7956e-06, 1.5625e+00, 6.4844e-01, + 1.5000e+00, 1.0625e+00, 4.8633e-01, -3.2227e-02, -3.9673e-03, + -9.9182e-04, 2.7222e-02, 1.3184e-01, -3.9101e-05, 6.4373e-05, + -4.3213e-06, -2.4033e-04, -5.0659e-03, -5.5176e-02, -3.1832e-10, + 3.2471e-02, -1.6809e-05, -2.2461e-02, -2.8320e-02, 8.9844e-01, + 1.2422e+00, 1.3438e+00, 1.3828e+00, 6.1328e-01, -2.0508e-02, + 9.5215e-03, 8.8215e-05, 1.2024e-02, 1.2283e-03, 9.0525e-07, + 1.7695e-07, 6.6895e-02, 9.3460e-05, 9.1903e-20, 1.0000e+00, + 1.1797e+00, 2.8564e-02, 2.2507e-04, 3.1494e-02, 1.1475e-01, + 1.1475e-02, 1.6719e+00, 9.8438e-01, 1.4609e+00, 1.5391e+00, + 2.9102e-01, 6.1328e-01, 5.2490e-02, -8.4229e-03, -1.1719e-02, + 1.7357e-04, -3.0396e-02, -1.8677e-02, -3.6316e-03, 4.1260e-02, + 1.3281e-01, -4.4584e-05, 1.1875e+00, -3.6478e-05, -9.1195e-06, + -9.1553e-05, -5.9372e-08, 1.1875e+00, 1.6328e+00, 1.5469e+00, + 1.5703e+00, 8.1250e-01, 1.9775e-02, -2.6001e-02, 3.2959e-03, + 3.4424e-02, 1.3770e-01, 1.4544e-05, 4.1723e-05, -3.1441e-06, + -4.1723e-06, -6.1512e-05, 1.2266e+00, -8.3674e-11, 3.0640e-02, + -5.6028e-06, 9.6484e-01, 1.2188e+00], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.self_attn.k_proj.weight': tensor([[-0.0034, 0.0308, 0.0071, ..., 0.0121, 0.0184, 0.0039], + [ 0.0065, 0.0037, 0.0056, ..., -0.0039, 0.0052, 0.0047], + [-0.0232, 0.0197, 0.0015, ..., 0.0042, 0.0145, 0.0150], + ..., + [ 0.0044, -0.0096, -0.0056, ..., 0.0079, 0.0087, 0.0111], + [ 0.0140, 0.0077, 0.0138, ..., -0.0079, -0.0084, 0.0176], + [ 0.0022, -0.0189, 0.0095, ..., -0.0082, 0.0088, 0.0299]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.k_norm.weight': tensor([ 1.4141e+00, 7.5781e-01, 4.7070e-01, 4.7461e-01, 1.3438e+00, + 1.5991e-02, 1.3977e-02, 8.8692e-05, 7.1716e-03, 1.5076e-02, + 3.2634e-06, -4.3656e-09, 1.1406e+00, 4.3869e-05, -3.5779e-18, + -3.7537e-03, -1.1206e-04, 9.2773e-02, 3.2196e-03, -3.2196e-03, + 4.0283e-03, -3.9339e-05, 1.0625e+00, 1.4609e+00, 4.9609e-01, + 1.2266e+00, 7.8125e-01, 2.9102e-01, 1.2451e-02, 5.8289e-03, + 1.0254e-02, 3.3379e-04, 8.3008e-03, 4.3213e-02, 1.8539e-03, + 9.0332e-03, 2.7710e-02, 1.7643e-05, 2.1729e-02, 2.1973e-03, + -8.8811e-06, -2.3395e-06, 1.6317e-06, 1.7188e+00, 1.1875e+00, + 1.5156e+00, 1.2734e+00, 4.8828e-01, -1.6357e-02, 6.4453e-02, + 3.1281e-03, -9.9487e-03, 4.7913e-03, -3.2902e-05, -1.7047e-05, + 3.9935e-06, -2.5177e-04, -5.0659e-03, 3.0640e-02, 2.7831e-10, + -3.9673e-03, -1.6212e-05, -3.0029e-02, -2.9907e-02, 6.8359e-01, + 1.1406e+00, 1.3750e+00, 1.3594e+00, 4.8242e-01, -1.9165e-02, + 6.1035e-03, -4.5300e-06, 3.5553e-03, -9.6893e-04, 1.1921e-06, + 1.8068e-07, 5.3406e-03, 5.8889e-05, 6.7424e-19, 1.0000e+00, + 1.1797e+00, 8.1055e-02, -4.8828e-04, -3.8300e-03, -3.6316e-03, + 1.1597e-02, 1.5859e+00, 1.2266e+00, 1.5703e+00, 1.5703e+00, + 8.7402e-02, 6.0547e-01, 2.3193e-02, 1.3977e-02, -7.9346e-03, + 1.0395e-04, 3.1738e-02, 6.0303e-02, 7.0801e-03, -1.0254e-02, + -2.5879e-02, -5.1022e-05, 1.1875e+00, 3.6478e-05, -2.4199e-05, + -9.1076e-05, -6.9384e-08, 1.0469e+00, 1.4766e+00, 1.4609e+00, + 1.6797e+00, 8.2031e-01, -2.5024e-02, 9.8633e-02, 3.6163e-03, + 1.9165e-02, -1.6602e-02, -2.7657e-05, 3.2425e-05, -2.9951e-06, + -2.0266e-06, 7.4387e-05, 1.2266e+00, -2.8081e-11, -1.0742e-02, + 5.3346e-06, 9.6484e-01, 1.2188e+00], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.self_attn.v_proj.weight': tensor([[-0.0051, 0.0036, -0.0222, ..., -0.0315, 0.0049, 0.0051], + [ 0.0030, 0.0017, -0.0138, ..., 0.0074, 0.0099, 0.0056], + [-0.0173, -0.0101, 0.0111, ..., 0.0151, -0.0081, -0.0086], + ..., + [ 0.0072, 0.0109, -0.0194, ..., -0.0356, 0.0075, 0.0027], + [ 0.0049, 0.0150, -0.0041, ..., 0.0060, 0.0060, -0.0049], + [ 0.0119, -0.0081, -0.0173, ..., 0.0171, -0.0041, 0.0049]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.output_proj.weight': tensor([[ 0.0085, 0.0261, -0.0267, ..., 0.0055, -0.0234, 0.0265], + [-0.0041, -0.0102, -0.0131, ..., 0.0220, 0.0256, -0.0087], + [ 0.0023, 0.0040, -0.0027, ..., -0.0017, 0.0041, 0.0084], + ..., + [ 0.0037, 0.0060, -0.0251, ..., 0.0096, 0.0140, -0.0101], + [-0.0144, 0.0079, 0.0062, ..., 0.0143, 0.0064, -0.0052], + [ 0.0088, -0.0034, -0.0133, ..., 0.0095, 0.0078, 0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_proj.weight': tensor([[-0.0112, 0.0037, -0.0019, ..., -0.0033, -0.0048, -0.0020], + [ 0.0121, 0.0047, 0.0027, ..., -0.0041, -0.0039, 0.0050], + [-0.0042, -0.0073, -0.0038, ..., -0.0074, 0.0123, 0.0027], + ..., + [-0.0063, -0.0048, -0.0037, ..., 0.0152, -0.0041, -0.0047], + [-0.0011, -0.0074, 0.0025, ..., 0.0134, -0.0055, -0.0047], + [ 0.0025, 0.0034, 0.0023, ..., -0.0037, 0.0043, 0.0037]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_norm.weight': tensor([-6.5613e-04, 1.3184e-01, 1.4526e-02, 6.0547e-01, 2.2339e-02, + 1.2634e-02, 8.5449e-03, 3.2806e-03, 7.0190e-03, 1.1279e-01, + 1.8677e-02, 5.0049e-02, 2.6367e-01, 1.3977e-02, 1.7090e-01, + 1.3245e-02, 5.3467e-02, 3.1494e-02, 1.6937e-03, 1.7578e-02, + 8.6060e-03, 1.8262e-01, 2.7344e-01, 1.8555e-01, 2.0752e-02, + 1.6016e-01, 5.5908e-02, 9.7656e-03, 4.0234e-01, 6.2866e-03, + 1.1292e-02, 5.0781e-01, 5.5664e-02, 3.3936e-02, 1.6113e-02, + 1.3123e-02, 6.7871e-02, 4.5508e-01, 5.3906e-01, 1.4954e-02, + 3.8910e-03, 1.0449e-01, 5.2490e-03, 5.3906e-01, 1.8188e-02, + 4.3213e-02, 1.0437e-02, 3.9258e-01, 7.2327e-03, 1.7944e-02, + 6.8359e-02, 1.8799e-02, 1.0437e-02, 5.0391e-01, -8.5831e-05, + 1.9897e-02, 5.5664e-02, 7.7148e-02, 9.0942e-03, 1.0132e-02, + 5.7983e-03, 9.6436e-03, 1.4099e-02, 3.3008e-01, 2.0264e-02, + 2.8534e-03, 2.6001e-02, 5.2979e-02, 6.2988e-02, 1.6235e-02, + 6.1951e-03, 6.8359e-01, -3.4094e-05, 6.1328e-01, 9.0332e-03, + 1.5198e-02, 2.2217e-02, 2.6953e-01, 3.2471e-02, 9.1171e-04, + 1.0645e-01, 9.7656e-03, 4.0771e-02, 1.1169e-02, 2.1362e-02, + 7.7209e-03, 2.9602e-03, 6.9336e-02, 1.9434e-01, 8.9111e-03, + 9.3994e-03, 1.5625e-02, 4.6875e-02, 2.8320e-02, 2.3193e-02, + 2.3560e-02, 1.4709e-02, 1.6846e-02, 6.2012e-02, 1.9287e-02, + 3.1250e-02, 7.2632e-03, 1.0803e-02, 5.4932e-02, 6.5234e-01, + 6.8848e-02, 2.6550e-03, 4.4531e-01, 1.0925e-02, 1.8799e-02, + 1.6479e-02, 1.5015e-02, 8.4043e-06, 8.9844e-02, 4.2969e-01, + 2.4414e-02, 1.7334e-02, 6.2109e-01, 1.1914e-01, 4.0039e-02, + 1.2146e-02, 7.1875e-01, 1.2891e-01, 1.8799e-02, 1.0925e-02, + 4.4250e-03, 2.0752e-02, 1.3046e-03], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_proj.weight': tensor([[-0.0046, -0.0034, 0.0103, ..., 0.0059, -0.0056, 0.0032], + [-0.0042, 0.0003, -0.0034, ..., 0.0005, -0.0081, 0.0029], + [-0.0002, -0.0016, -0.0085, ..., 0.0123, -0.0168, -0.0072], + ..., + [ 0.0040, 0.0046, -0.0043, ..., 0.0052, 0.0020, -0.0061], + [ 0.0093, 0.0029, -0.0031, ..., -0.0035, 0.0093, 0.0023], + [-0.0056, 0.0007, -0.0092, ..., -0.0081, -0.0060, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_norm.weight': tensor([ 6.8665e-04, 1.3184e-01, 1.4038e-02, 6.0156e-01, 2.1851e-02, + 1.2573e-02, 8.7280e-03, 3.2654e-03, 7.4158e-03, 1.1279e-01, + 1.8677e-02, 4.8584e-02, 2.6367e-01, 1.3855e-02, 1.7090e-01, + 1.3916e-02, 5.2734e-02, 3.1738e-02, 1.6327e-03, 1.7456e-02, + 8.9111e-03, 1.8359e-01, 2.7539e-01, 1.8555e-01, 2.0508e-02, + 1.6211e-01, 5.4443e-02, 7.9956e-03, 4.0430e-01, 6.2561e-03, + 1.1536e-02, 5.0781e-01, 5.5664e-02, 3.3936e-02, 1.6479e-02, + 1.2451e-02, 6.8359e-02, 4.5508e-01, 5.3906e-01, 1.4954e-02, + 3.8910e-03, 1.0449e-01, 5.2185e-03, 5.3906e-01, 1.8433e-02, + 4.3213e-02, 1.0803e-02, 3.9258e-01, 7.3853e-03, 1.7822e-02, + 6.8359e-02, 1.9043e-02, 1.0376e-02, 5.0391e-01, 1.2589e-04, + 1.9897e-02, 5.5664e-02, 7.7148e-02, 9.0942e-03, 8.3008e-03, + 5.7373e-03, 9.6436e-03, 1.4404e-02, 3.3008e-01, 1.9775e-02, + 2.9755e-03, 2.5513e-02, 5.2979e-02, 6.2988e-02, 1.6235e-02, + 6.5918e-03, 6.7969e-01, -3.3855e-05, 6.1328e-01, 9.2163e-03, + -1.4771e-02, 2.2217e-02, 2.6953e-01, 3.3203e-02, 9.1934e-04, + 1.0693e-01, 9.6436e-03, 4.0039e-02, 1.0986e-02, 2.0874e-02, + 7.7820e-03, 3.9062e-03, 6.9824e-02, 1.9434e-01, 8.9111e-03, + 9.5825e-03, 1.4709e-02, 4.7119e-02, 2.8442e-02, 2.4048e-02, + 2.3682e-02, 1.4709e-02, 1.6968e-02, 6.2012e-02, 1.9165e-02, + 3.1250e-02, 7.4768e-03, 1.1230e-02, 5.5420e-02, 6.4844e-01, + 6.8848e-02, 2.6855e-03, 4.4727e-01, 1.0620e-02, 1.8677e-02, + 1.6968e-02, 1.5015e-02, 1.1623e-05, 8.9355e-02, 4.2969e-01, + 2.4170e-02, 1.7212e-02, 6.2109e-01, 1.1914e-01, 4.0283e-02, + 1.2085e-02, 7.2266e-01, 1.2793e-01, 1.7822e-02, 1.0681e-02, + 4.3640e-03, 2.0630e-02, 7.3624e-04], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.v_proj.weight': tensor([[ 0.0170, -0.0073, -0.0155, ..., 0.0033, -0.0150, -0.0077], + [-0.0186, -0.0008, -0.0024, ..., -0.0024, -0.0233, 0.0194], + [-0.0188, -0.0067, -0.0031, ..., -0.0038, 0.0283, 0.0312], + ..., + [-0.0130, -0.0038, -0.0086, ..., -0.0001, 0.0154, 0.0029], + [ 0.0229, 0.0045, 0.0222, ..., 0.0052, 0.0056, 0.0077], + [ 0.0139, 0.0012, 0.0033, ..., 0.0089, -0.0164, -0.0210]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0003, 0.0027, ..., -0.0012, 0.0021, 0.0156], + [ 0.0109, -0.0014, 0.0027, ..., 0.0042, 0.0002, -0.0014], + [ 0.0042, 0.0008, -0.0021, ..., -0.0047, -0.0039, -0.0054], + ..., + [-0.0178, -0.0018, 0.0037, ..., 0.0044, -0.0113, 0.0034], + [ 0.0046, -0.0100, 0.0056, ..., -0.0310, 0.0078, -0.0223], + [ 0.0013, -0.0036, 0.0085, ..., 0.0297, -0.0044, 0.0217]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0064, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [-0.0015, 0.0150, -0.0159, ..., -0.0041, -0.0055, 0.0679], + ..., + [-0.0208, 0.0130, -0.0273, ..., 0.0308, -0.0072, -0.0068], + [-0.0337, -0.0025, 0.0054, ..., -0.0040, 0.0172, 0.0231], + [-0.0255, 0.0040, 0.0030, ..., 0.0408, 0.0430, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0031], + [ 0.0198, -0.0208, 0.0130, ..., 0.0089, -0.0337, -0.0055], + [ 0.0302, 0.0101, -0.0023, ..., -0.0304, 0.0199, 0.0024], + ..., + [ 0.0171, -0.0275, -0.0075, ..., 0.0231, 0.0182, -0.0039], + [ 0.0221, 0.0388, 0.0056, ..., 0.0042, -0.0488, -0.0361], + [-0.0005, 0.0120, -0.0051, ..., 0.0037, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img.weight': tensor([[ 0.0273, -0.0041, 0.0118, ..., 0.0012, 0.0347, 0.0615], + [-0.0258, 0.0042, -0.0197, ..., 0.0139, -0.0047, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0097, 0.0132, 0.0184, ..., 0.0110, 0.0087, 0.0515], + [-0.0006, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.mlp.layer1.weight': tensor([[ 0.0017, -0.0165, 0.0046, ..., 0.0017, -0.0049, -0.0064], + [-0.0162, 0.0078, -0.0029, ..., 0.0124, 0.0065, -0.0203], + [-0.0122, -0.0107, 0.0084, ..., -0.0070, -0.0010, -0.0610], + ..., + [-0.0042, 0.0038, 0.0038, ..., 0.0156, 0.0085, -0.0038], + [-0.0093, 0.0140, -0.0031, ..., -0.0092, 0.0054, -0.0396], + [-0.0037, 0.0162, -0.0016, ..., 0.0114, 0.0172, -0.0091]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.mlp.layer2.weight': tensor([[ 0.0040, -0.0082, -0.0236, ..., 0.0062, -0.0112, -0.0124], + [-0.0309, -0.0120, -0.0055, ..., -0.0011, -0.0064, -0.0074], + [ 0.0049, -0.0071, -0.0030, ..., 0.0031, -0.0033, 0.0087], + ..., + [-0.0028, -0.0079, -0.0432, ..., -0.0063, -0.0098, -0.0110], + [-0.0104, 0.0050, 0.0120, ..., 0.0024, 0.0128, -0.0052], + [-0.0078, -0.0063, -0.0037, ..., -0.0096, 0.0097, -0.0388]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0396, 0.0398, 0.0381, ..., -0.0042, -0.0042, -0.0044], + [-0.0383, -0.0369, -0.0359, ..., 0.0043, 0.0048, 0.0047], + [-0.0400, -0.0393, -0.0383, ..., 0.0045, 0.0049, 0.0049], + ..., + [-0.0422, -0.0408, -0.0400, ..., 0.0022, 0.0022, 0.0024], + [-0.0415, -0.0408, -0.0398, ..., 0.0021, 0.0022, 0.0022], + [-0.0039, -0.0039, -0.0039, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_self_attn.2.weight': tensor([[ 3.8757e-03, -3.4637e-03, -3.5248e-03, ..., 2.0504e-05, + -5.6744e-05, 5.4016e-03], + [ 3.3722e-03, -4.0283e-03, -3.2654e-03, ..., -4.7302e-03, + -4.6997e-03, -1.7776e-03], + [ 4.8218e-03, -6.3171e-03, -5.7678e-03, ..., -3.7842e-03, + -3.8147e-03, 1.3123e-03], + ..., + [ 3.9062e-02, -4.1992e-02, -4.2480e-02, ..., -2.0630e-02, + -1.7334e-02, 1.6235e-02], + [ 3.9978e-03, -5.7983e-03, -5.8899e-03, ..., 3.8910e-03, + 3.8910e-03, 5.5847e-03], + [-1.4038e-02, 1.5564e-02, 1.4099e-02, ..., 1.7090e-02, + 1.6968e-02, 1.4526e-02]], dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9570e-02, 5.8838e-02, 5.7617e-02, ..., -1.1015e-04, + -3.9339e-05, -1.9169e-04], + [ 6.3965e-02, 6.2012e-02, 5.9326e-02, ..., 3.4904e-04, + 4.6158e-04, 4.9210e-04], + [-4.8340e-02, -4.6387e-02, -4.5654e-02, ..., -2.6464e-05, + 2.8014e-05, -8.0109e-05], + ..., + [ 3.1494e-02, 3.0518e-02, 2.9785e-02, ..., 1.1206e-04, + 4.5061e-05, 7.3433e-05], + [ 6.1035e-02, 5.9326e-02, 5.8594e-02, ..., 5.2643e-04, + 5.6076e-04, 5.5313e-04], + [-4.7852e-02, -4.6143e-02, -4.5166e-02, ..., -3.2234e-04, + -3.6430e-04, -3.2806e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0149, 0.0073, -0.0080, ..., 0.0018, -0.0047, -0.0041], + [-0.0178, 0.0121, 0.0144, ..., -0.0096, -0.0124, 0.0083], + [ 0.0011, 0.0030, -0.0011, ..., 0.0009, 0.0003, -0.0002], + ..., + [ 0.0260, 0.0184, -0.0149, ..., 0.0078, 0.0080, 0.0037], + [ 0.0075, 0.0334, -0.0057, ..., 0.0076, 0.0325, -0.0178], + [-0.0044, 0.0203, 0.0051, ..., -0.0114, -0.0110, 0.0240]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_mlp.1.weight': tensor([[-0.0986, -0.0967, -0.0942, ..., 0.0031, 0.0034, 0.0033], + [-0.0371, -0.0352, -0.0339, ..., 0.0008, 0.0008, 0.0007], + [ 0.0566, 0.0549, 0.0544, ..., -0.0017, -0.0017, -0.0017], + ..., + [ 0.0364, 0.0354, 0.0347, ..., -0.0018, -0.0018, -0.0016], + [ 0.0466, 0.0444, 0.0430, ..., -0.0010, -0.0009, -0.0010], + [-0.0518, -0.0505, -0.0496, ..., 0.0022, 0.0022, 0.0021]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_mlp.2.weight': tensor([[-0.0228, -0.0165, 0.0110, ..., -0.0018, 0.0181, -0.0115], + [ 0.0017, -0.0033, -0.0016, ..., -0.0016, 0.0030, 0.0016], + [-0.0005, -0.0007, -0.0008, ..., -0.0010, 0.0007, 0.0009], + ..., + [ 0.0454, -0.0031, -0.0184, ..., -0.0126, 0.0032, 0.0193], + [ 0.0459, -0.0288, -0.0371, ..., -0.0337, 0.0286, 0.0366], + [ 0.0128, -0.0366, -0.0251, ..., -0.0284, 0.0374, 0.0240]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.after_proj.weight': tensor([[-0.0054, -0.0084, -0.0073, ..., 0.0059, 0.0018, -0.0046], + [ 0.0066, 0.0057, 0.0067, ..., 0.0050, 0.0052, 0.0131], + [-0.0096, -0.0041, -0.0051, ..., -0.0023, -0.0037, 0.0110], + ..., + [-0.0015, -0.0067, -0.0140, ..., -0.0055, -0.0112, 0.0187], + [-0.0049, -0.0150, -0.0076, ..., 0.0105, -0.0118, -0.0096], + [-0.0036, 0.0160, 0.0104, ..., -0.0038, 0.0221, -0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.after_proj.bias': tensor([-3.4790e-03, 2.1820e-03, 1.1139e-03, ..., 4.6387e-03, + 9.3937e-05, -4.8637e-04], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_proj.weight': tensor([[-0.0110, 0.0153, 0.0214, ..., 0.0090, 0.0057, -0.0190], + [ 0.0203, -0.0014, 0.0074, ..., 0.0058, -0.0027, 0.0075], + [-0.0127, -0.0063, 0.0129, ..., -0.0188, -0.0093, -0.0134], + ..., + [ 0.0065, 0.0038, 0.0002, ..., 0.0057, 0.0030, 0.0051], + [ 0.0098, -0.0222, -0.0029, ..., -0.0398, -0.0059, -0.0120], + [-0.0065, -0.0048, -0.0052, ..., 0.0099, 0.0019, 0.0074]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_norm.weight': tensor([ 1.3672e+00, 1.8047e+00, 1.3828e+00, 1.5859e+00, 2.2852e-01, + 5.6396e-02, 6.5308e-03, 4.2915e-06, -1.7694e-33, 3.5477e-04, + 3.9062e-02, 1.2695e-01, 1.0620e-02, 3.2959e-02, -1.1719e-02, + 4.9744e-03, 6.3965e-02, 9.5703e-01, 8.8281e-01, 1.0376e-03, + 8.9169e-05, 1.2734e+00, 2.2188e+00, 1.9609e+00, 1.0469e+00, + 1.1094e+00, 1.5312e+00, 8.9264e-04, -1.6689e-04, -5.9128e-04, + 1.1292e-02, 1.7944e-02, 9.9487e-03, 1.0156e+00, 9.1553e-03, + 1.0156e+00, 7.3242e-04, -6.8665e-05, 7.4768e-03, -1.2024e-02, + 1.9670e-06, -5.9605e-06, 6.8665e-05, 2.3125e+00, 1.5000e+00, + 1.5781e+00, 1.6016e+00, 1.7578e+00, 1.1523e-01, 7.4768e-04, + 2.4048e-02, 2.4707e-01, 1.1172e+00, 1.3281e-01, 1.4038e-03, + 2.1973e-01, 3.0327e-04, 1.2305e-01, 6.1035e-05, -4.4861e-03, + -7.4463e-03, -5.6152e-03, -6.3477e-02, 1.1328e+00, 1.9609e+00, + 8.3984e-01, 1.1094e+00, 1.3281e+00, 7.7637e-02, -2.1484e-02, + 1.2573e-02, 2.5511e-05, -2.9491e-33, -1.1597e-03, 5.2795e-03, + 1.0559e-02, 1.4496e-03, -3.5156e-02, -2.8442e-02, 1.7944e-02, + 8.8867e-02, 3.6133e-02, 1.0254e-02, 6.1951e-03, -3.7994e-03, + 2.8687e-03, 1.0156e+00, 1.1953e+00, 1.6094e+00, 1.5703e+00, + 9.3750e-01, -7.4768e-03, 5.8365e-04, -2.9297e-03, 3.1281e-03, + -6.8665e-03, 2.1729e-02, 5.6152e-02, 1.7090e-02, 4.4678e-02, + -2.8229e-03, -1.3199e-03, 1.5234e-01, 4.2439e-05, 5.4550e-04, + -8.8811e-06, 2.8534e-03, 8.0469e-01, 1.7734e+00, 1.5547e+00, + 1.1016e+00, 1.2031e+00, 6.2988e-02, 6.5994e-04, 4.1992e-02, + 1.2188e+00, 1.1035e-01, 1.1641e+00, -2.0790e-04, 1.3794e-02, + 1.1475e-02, 5.6641e-02, 2.0313e-04, -2.2949e-02, -5.6028e-05, + 1.2695e-02, 1.0234e+00, -9.7656e-03], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.self_attn.k_proj.weight': tensor([[-0.0220, -0.0096, -0.0002, ..., -0.0080, 0.0030, -0.0188], + [ 0.0066, -0.0347, -0.0034, ..., 0.0116, 0.0139, 0.0144], + [-0.0121, 0.0195, -0.0061, ..., -0.0107, -0.0099, 0.0030], + ..., + [ 0.0010, 0.0106, -0.0084, ..., -0.0008, 0.0042, 0.0038], + [ 0.0222, -0.0081, 0.0102, ..., -0.0147, -0.0076, 0.0221], + [-0.0123, -0.0031, -0.0052, ..., 0.0056, 0.0062, 0.0053]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.k_norm.weight': tensor([ 9.6094e-01, 1.7266e+00, 1.3984e+00, 1.4844e+00, 2.3535e-01, + 6.2500e-02, -1.8845e-03, 1.1861e-05, -1.5167e-33, 2.5177e-04, + 3.9551e-02, 1.2793e-01, -1.1780e-02, 5.1025e-02, -1.1902e-02, + -4.7607e-03, 1.0925e-02, 9.5703e-01, 8.8281e-01, -9.4986e-04, + 2.2316e-04, 1.2734e+00, 1.8828e+00, 1.7734e+00, 4.3945e-02, + 5.3906e-01, 1.5469e+00, -6.9580e-03, -3.4714e-04, 4.4823e-04, + -1.1597e-02, 1.7822e-02, -6.1340e-03, 1.0078e+00, -7.6599e-03, + 1.0078e+00, 7.4768e-04, -1.0681e-04, -9.5215e-03, -1.2146e-02, + 2.9206e-06, -7.8082e-06, 6.4373e-05, 1.8672e+00, 1.0234e+00, + 1.5391e+00, 1.6641e+00, 1.7109e+00, 7.5684e-02, 8.0872e-04, + -3.9368e-03, -6.7444e-03, 1.1172e+00, 1.7188e-01, 1.3351e-03, + 2.1973e-01, -8.8501e-04, 1.2305e-01, -1.3173e-05, 7.1289e-02, + -7.4768e-03, 2.4414e-02, -7.1777e-02, 1.1328e+00, 2.0781e+00, + 1.2109e+00, 9.2969e-01, 1.3828e+00, -1.7944e-02, -1.0254e-02, + -1.6968e-02, -2.2888e-05, 4.3815e-33, 1.3275e-03, 1.7456e-02, + 4.6143e-02, 6.2561e-04, 4.9561e-02, -3.9795e-02, -1.7944e-02, + -5.6152e-02, 2.8809e-02, 1.3062e-02, 6.1340e-03, 3.9673e-03, + -1.1215e-03, 1.5234e+00, 1.5625e+00, 1.6953e+00, 1.6172e+00, + 5.4297e-01, 5.5542e-03, -4.3297e-04, -2.9907e-03, 9.4223e-04, + 7.1411e-03, -3.7384e-03, 5.9509e-03, -1.6235e-02, 8.3008e-03, + -2.8076e-03, -1.3275e-03, 1.5234e-01, 3.1233e-05, 5.6839e-04, + -1.0788e-05, 2.8534e-03, 1.6250e+00, 1.9844e+00, 1.5469e+00, + 5.5859e-01, 1.2969e+00, -4.3030e-03, 2.6512e-04, -3.1738e-02, + 1.2188e+00, 1.5442e-02, 1.1641e+00, -2.6512e-04, 1.6724e-02, + 1.1719e-02, 5.5908e-02, 2.0313e-04, 5.7617e-02, -1.0014e-04, + 2.2949e-02, 1.0234e+00, 3.2959e-03], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.self_attn.v_proj.weight': tensor([[ 0.0020, -0.0152, 0.0064, ..., 0.0461, -0.0211, 0.0038], + [ 0.0127, 0.0267, -0.0182, ..., 0.0046, 0.0138, -0.0134], + [-0.0003, 0.0297, -0.0025, ..., -0.0084, 0.0220, 0.0025], + ..., + [ 0.0086, 0.0102, -0.0201, ..., 0.0080, 0.0078, 0.0047], + [-0.0018, 0.0332, 0.0127, ..., 0.0145, 0.0266, -0.0056], + [-0.0086, -0.0042, 0.0227, ..., 0.0187, 0.0172, -0.0100]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.output_proj.weight': tensor([[ 0.0066, 0.0055, 0.0227, ..., 0.0034, 0.0201, -0.0080], + [ 0.0229, -0.0079, 0.0054, ..., -0.0186, 0.0041, -0.0114], + [ 0.0059, 0.0130, -0.0233, ..., -0.0077, -0.0072, 0.0134], + ..., + [-0.0129, -0.0464, -0.0078, ..., -0.0089, -0.0006, -0.0193], + [ 0.0032, 0.0084, -0.0201, ..., 0.0087, 0.0236, 0.0068], + [-0.0566, -0.0126, 0.0322, ..., 0.0216, -0.0221, -0.0166]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_proj.weight': tensor([[ 0.0098, -0.0079, 0.0138, ..., 0.0038, 0.0074, 0.0201], + [ 0.0078, -0.0058, -0.0052, ..., -0.0090, -0.0052, 0.0119], + [-0.0033, 0.0165, -0.0048, ..., 0.0044, -0.0076, -0.0115], + ..., + [ 0.0093, 0.0032, -0.0132, ..., 0.0160, 0.0046, -0.0068], + [-0.0110, -0.0153, -0.0032, ..., -0.0090, 0.0082, 0.0144], + [ 0.0095, -0.0047, -0.0076, ..., 0.0019, 0.0171, -0.0103]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_norm.weight': tensor([0.1064, 0.0159, 0.4277, 0.0698, 0.1680, 0.0361, 0.1484, 0.1357, 0.5938, + 0.0364, 0.0618, 0.0195, 0.0491, 0.0767, 0.0566, 0.1963, 0.4844, 0.0320, + 0.7461, 0.0352, 0.0437, 0.1689, 0.0178, 0.0801, 0.0481, 0.4082, 0.5508, + 0.2559, 0.0771, 0.0535, 0.0403, 0.2158, 0.6680, 0.2441, 0.0518, 0.0181, + 0.0105, 0.1099, 0.0156, 0.0476, 0.0493, 0.1289, 0.0182, 0.0400, 0.2041, + 0.0532, 0.3555, 0.0205, 0.0500, 0.0139, 0.0310, 0.3848, 0.5391, 0.3438, + 0.0903, 0.6484, 0.5977, 0.0400, 0.1235, 0.0322, 0.0854, 0.5586, 0.0102, + 0.5508, 0.1719, 0.3926, 0.0825, 0.0508, 0.0625, 0.4688, 0.0242, 0.5234, + 0.0214, 0.0356, 0.1338, 0.0874, 0.0287, 0.0889, 0.6367, 0.0352, 0.0889, + 0.0825, 0.0435, 0.0391, 0.4219, 0.0640, 0.0625, 0.0986, 0.0352, 0.1426, + 0.0459, 0.1758, 0.0537, 0.0588, 0.0295, 0.1221, 0.0131, 0.3613, 0.5820, + 0.0342, 0.0562, 0.2949, 0.0208, 0.1211, 0.0442, 0.2314, 0.0228, 0.6328, + 0.5859, 0.0327, 0.0120, 0.7695, 0.0192, 0.0137, 0.4941, 0.2070, 0.2754, + 0.7500, 0.0410, 0.0757, 0.0776, 0.5781, 0.6406, 0.0125, 0.0272, 0.2021, + 0.0249, 0.0205], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_proj.weight': tensor([[-8.7280e-03, 1.8158e-03, 7.0496e-03, ..., -5.6458e-04, + 2.1973e-02, -2.3346e-03], + [ 7.2327e-03, -6.7711e-05, -4.7302e-03, ..., -2.6398e-03, + -2.5330e-03, 1.0986e-02], + [-1.3123e-02, 7.6904e-03, -1.5747e-02, ..., 2.2888e-03, + 6.5002e-03, 5.9509e-03], + ..., + [ 2.8839e-03, 3.1128e-03, -2.5330e-03, ..., 2.6855e-03, + -7.0801e-03, 1.2390e-02], + [-1.1780e-02, -1.8539e-03, -1.4465e-02, ..., 5.3406e-03, + -9.0332e-03, 2.7466e-03], + [ 1.0010e-02, -1.4038e-03, 2.1820e-03, ..., -1.3855e-02, + 5.8594e-03, -1.8997e-03]], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_norm.weight': tensor([0.1064, 0.0157, 0.4277, 0.0698, 0.1680, 0.0364, 0.1484, 0.1348, 0.5938, + 0.0369, 0.0623, 0.0195, 0.0491, 0.0767, 0.0569, 0.1982, 0.4844, 0.0320, + 0.7422, 0.0349, 0.0435, 0.1680, 0.0178, 0.0796, 0.0481, 0.4043, 0.5508, + 0.2539, 0.0771, 0.0532, 0.0413, 0.2178, 0.6680, 0.2441, 0.0525, 0.0181, + 0.0102, 0.1099, 0.0159, 0.0466, 0.0488, 0.1299, 0.0183, 0.0403, 0.2041, + 0.0532, 0.3555, 0.0206, 0.0498, 0.0137, 0.0315, 0.3848, 0.5352, 0.3438, + 0.0903, 0.6484, 0.6016, 0.0398, 0.1235, 0.0325, 0.0854, 0.5586, 0.0106, + 0.5508, 0.1709, 0.3926, 0.0825, 0.0505, 0.0625, 0.4707, 0.0240, 0.5234, + 0.0212, 0.0356, 0.1338, 0.0869, 0.0286, 0.0898, 0.6367, 0.0347, 0.0879, + 0.0830, 0.0435, 0.0391, 0.4219, 0.0645, 0.0618, 0.0977, 0.0354, 0.1445, + 0.0469, 0.1748, 0.0537, 0.0596, 0.0298, 0.1216, 0.0131, 0.3613, 0.5820, + 0.0337, 0.0569, 0.2949, 0.0206, 0.1211, 0.0442, 0.2314, 0.0225, 0.6328, + 0.5859, 0.0325, 0.0119, 0.7695, 0.0187, 0.0140, 0.4961, 0.2070, 0.2754, + 0.7461, 0.0403, 0.0732, 0.0771, 0.5781, 0.6406, 0.0123, 0.0272, 0.2021, + 0.0261, 0.0210], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.v_proj.weight': tensor([[ 0.0135, -0.0027, -0.0303, ..., -0.0010, -0.0011, -0.0116], + [-0.0089, 0.0164, -0.0017, ..., 0.0066, -0.0045, -0.0234], + [ 0.0212, 0.0015, -0.0043, ..., 0.0004, -0.0052, -0.0410], + ..., + [-0.0072, 0.0032, 0.0109, ..., -0.0109, -0.0275, 0.0041], + [-0.0396, -0.0017, -0.0281, ..., -0.0083, 0.0090, -0.0190], + [-0.0095, -0.0029, 0.0029, ..., -0.0099, 0.0098, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.output_proj.weight': tensor([[ 0.0021, 0.0048, -0.0103, ..., -0.0038, 0.0046, 0.0111], + [-0.0361, 0.0028, -0.0017, ..., 0.0057, -0.0110, -0.0220], + [ 0.0036, -0.0078, -0.0037, ..., 0.0044, 0.0173, -0.0050], + ..., + [ 0.0079, -0.0135, 0.0210, ..., 0.0182, -0.0076, -0.0128], + [-0.0109, -0.0053, 0.0071, ..., 0.0124, -0.0144, 0.0092], + [-0.0041, -0.0201, 0.0154, ..., -0.0038, 0.0083, 0.0167]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0050, -0.0065, ..., 0.0266, -0.0278, 0.0078], + ..., + [-0.0217, -0.0410, 0.0488, ..., 0.0039, -0.0022, -0.0255], + [-0.0032, 0.0066, 0.0034, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0055, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, -0.0019, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0039, 0.0293, ..., 0.0065, 0.0089, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0140, 0.0126, -0.0093], + ..., + [-0.0044, 0.0320, -0.0233, ..., -0.0067, 0.0038, 0.0187], + [ 0.0136, -0.0352, 0.0267, ..., -0.0305, -0.0039, 0.0005], + [-0.0205, 0.0143, 0.0008, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0014, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0058, -0.0537, -0.0282, ..., -0.0109, -0.0146, -0.0023], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, 0.0029, 0.0015], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0024, ..., -0.0103, 0.0073, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.mlp.layer1.weight': tensor([[ 0.0123, 0.0036, 0.0047, ..., 0.0121, -0.0084, -0.0044], + [-0.0125, 0.0001, 0.0205, ..., 0.0079, -0.0267, -0.0065], + [ 0.0048, 0.0181, -0.0098, ..., 0.0228, -0.0131, 0.0217], + ..., + [-0.0101, 0.0042, 0.0571, ..., 0.0067, 0.0117, -0.0193], + [ 0.0087, -0.0047, -0.0078, ..., -0.0101, -0.0037, -0.0166], + [ 0.0046, 0.0154, -0.0159, ..., -0.0304, -0.0219, -0.0162]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.mlp.layer2.weight': tensor([[ 0.0048, 0.0209, -0.0045, ..., -0.0162, -0.0251, 0.0036], + [ 0.0038, 0.0232, 0.0094, ..., -0.0051, 0.0177, 0.0135], + [ 0.0050, -0.0222, 0.0045, ..., -0.0055, -0.0045, 0.0055], + ..., + [-0.0053, 0.0055, 0.0253, ..., -0.0049, -0.0100, 0.0018], + [ 0.0039, 0.0140, 0.0219, ..., -0.0047, 0.0172, 0.0111], + [-0.0066, 0.0059, -0.0039, ..., -0.0128, 0.0103, 0.0061]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0469, -0.0464, -0.0449, ..., 0.0010, 0.0008, 0.0009], + [-0.0386, -0.0381, -0.0371, ..., 0.0027, 0.0027, 0.0027], + [-0.0400, -0.0391, -0.0376, ..., 0.0022, 0.0023, 0.0023], + ..., + [ 0.0381, 0.0371, 0.0359, ..., -0.0036, -0.0037, -0.0036], + [ 0.0327, 0.0317, 0.0306, ..., -0.0031, -0.0030, -0.0030], + [-0.0386, -0.0376, -0.0371, ..., 0.0022, 0.0024, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0152, 0.0145, 0.0137, ..., 0.0220, 0.0248, 0.0148], + [ 0.0016, -0.0010, 0.0036, ..., 0.0025, 0.0036, 0.0003], + [-0.0003, -0.0022, -0.0009, ..., 0.0038, 0.0041, -0.0025], + ..., + [-0.0250, -0.0747, -0.0610, ..., 0.0364, 0.0369, -0.0679], + [ 0.0084, 0.0085, 0.0068, ..., 0.0199, 0.0176, 0.0087], + [ 0.0142, 0.0089, 0.0103, ..., 0.0192, 0.0216, 0.0101]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_cross_attn.1.weight': tensor([[-7.9102e-02, -7.6172e-02, -7.4707e-02, ..., -1.0252e-04, + -1.1730e-04, -1.0729e-04], + [ 3.0029e-02, 2.9297e-02, 2.8809e-02, ..., 9.4995e-07, + 2.6345e-05, 1.4687e-04], + [ 2.8442e-02, 2.7710e-02, 2.6489e-02, ..., 2.7466e-04, + 2.9755e-04, 2.0504e-04], + ..., + [ 6.4453e-02, 6.2988e-02, 6.2012e-02, ..., -1.8978e-04, + -1.8406e-04, -1.1444e-04], + [ 6.6895e-02, 6.4453e-02, 6.2988e-02, ..., -1.7643e-04, + -1.1778e-04, -1.6403e-04], + [-3.6133e-02, -3.5156e-02, -3.4424e-02, ..., -5.2929e-05, + 9.9659e-05, 1.2112e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0038, -0.0112, -0.0028, ..., 0.0201, 0.0288, 0.0041], + [ 0.0071, -0.0090, 0.0116, ..., 0.0005, 0.0060, -0.0019], + [ 0.0011, -0.0026, 0.0036, ..., -0.0082, -0.0055, -0.0093], + ..., + [-0.0141, 0.0030, 0.0330, ..., 0.0052, 0.0139, 0.0052], + [-0.0225, 0.0114, 0.0369, ..., 0.0081, 0.0171, -0.0190], + [ 0.0266, -0.0216, 0.0052, ..., -0.0014, 0.0077, 0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_mlp.1.weight': tensor([[-8.9844e-02, -8.6426e-02, -8.3984e-02, ..., 1.8787e-04, + 3.0899e-04, 2.5368e-04], + [ 2.2461e-02, 2.0264e-02, 1.9409e-02, ..., 1.8787e-04, + 2.6894e-04, 2.6321e-04], + [-9.2285e-02, -9.0332e-02, -8.9844e-02, ..., -4.0245e-04, + -3.2806e-04, -3.6621e-04], + ..., + [-8.2031e-02, -8.0566e-02, -7.9102e-02, ..., 2.1076e-04, + 3.1471e-04, 2.1172e-04], + [-6.1279e-02, -5.9570e-02, -5.8594e-02, ..., 2.2888e-04, + 9.2506e-05, 1.0252e-04], + [ 7.9102e-02, 7.7637e-02, 7.6660e-02, ..., -1.7738e-04, + -2.0599e-04, -1.6880e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_mlp.2.weight': tensor([[ 0.0023, 0.0038, 0.0038, ..., -0.0055, -0.0086, 0.0069], + [-0.0018, -0.0021, 0.0057, ..., -0.0016, 0.0037, 0.0015], + [-0.0010, 0.0045, -0.0019, ..., -0.0009, -0.0012, 0.0011], + ..., + [-0.0413, 0.0198, -0.0593, ..., -0.0295, -0.0106, 0.0225], + [ 0.0091, -0.0100, 0.0258, ..., 0.0067, 0.0037, -0.0063], + [ 0.0154, 0.0053, 0.0056, ..., 0.0135, 0.0070, -0.0107]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.after_proj.weight': tensor([[-0.0056, 0.0069, 0.0043, ..., 0.0067, 0.0135, -0.0127], + [-0.0050, -0.0069, 0.0068, ..., -0.0057, 0.0049, -0.0186], + [ 0.0104, -0.0056, 0.0124, ..., 0.0151, 0.0187, 0.0063], + ..., + [-0.0077, -0.0034, 0.0146, ..., 0.0031, -0.0150, -0.0096], + [-0.0122, -0.0050, -0.0278, ..., -0.0188, 0.0081, -0.0032], + [-0.0042, 0.0045, 0.0044, ..., -0.0145, 0.0129, 0.0003]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.after_proj.bias': tensor([ 0.0081, -0.0087, -0.0045, ..., -0.0044, -0.0139, -0.0038], + dtype=torch.bfloat16)} +[06-26 07:35:41|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:41|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 07:35:41|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 0 checkpoint. +[06-26 07:35:41|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/blur(ba2f44f2-c726-4fe7-949f-597069d9b91c) +[06-26 07:35:41|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 1 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/eb5325b77d358944da58a690157dd2b8071bbf85/general/blur/ba2f44f2-c726-4fe7-949f-597069d9b91c_ema_bf16.pt +[06-26 07:35:42|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0142, 0.0030, 0.0065], + [-0.0017, 0.0060, -0.0327, ..., -0.0070, 0.0009, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0135, -0.0033, 0.0283], + ..., + [-0.0031, 0.0125, 0.0248, ..., 0.0437, -0.0069, -0.0192], + [ 0.0277, 0.0146, 0.0040, ..., -0.0038, -0.0090, 0.0146], + [ 0.0437, 0.0366, 0.0085, ..., 0.0038, -0.0129, 0.0116]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0045, -0.0310, 0.0300, ..., -0.0068, -0.0048, -0.0339], + [-0.0103, -0.0048, 0.0115, ..., 0.0048, 0.0092, 0.0520], + [-0.0061, 0.0132, 0.0398, ..., -0.0096, -0.0179, -0.0061], + ..., + [-0.0046, 0.0175, -0.0095, ..., 0.0225, 0.0010, 0.0030], + [-0.0209, -0.0081, 0.0087, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0055, 0.0354, -0.0081, ..., 0.0118, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0011, ..., 0.0376, 0.0532, 0.0014], + [ 0.0175, -0.0347, -0.0013, ..., -0.0057, 0.0050, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0154, 0.0262, -0.0112], + [-0.0001, 0.0244, 0.0356, ..., 0.0019, 0.0170, 0.0119], + [ 0.0486, 0.0175, 0.0042, ..., -0.0020, -0.0023, -0.0021]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0141, 0.0569, 0.0115, ..., -0.0063, -0.0337, 0.0076], + [ 0.0491, 0.0413, 0.0251, ..., -0.0053, 0.0393, 0.0047], + [-0.0162, -0.0033, 0.0134, ..., -0.0198, -0.0244, 0.0469], + ..., + [ 0.0014, 0.0261, 0.0022, ..., -0.0041, 0.0347, -0.0033], + [-0.0254, 0.0068, -0.0016, ..., 0.0096, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0122, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0033, -0.0270, -0.0166, ..., 0.0118, -0.0298, 0.0532], + [-0.0031, -0.0035, -0.0128, ..., -0.0161, 0.0149, 0.0308], + [ 0.0144, 0.0086, 0.0146, ..., -0.0118, -0.0410, 0.0354], + ..., + [ 0.0062, -0.0630, -0.0034, ..., 0.0072, -0.0183, 0.0217], + [ 0.0038, 0.0036, 0.0258, ..., -0.0486, -0.0349, -0.0036], + [ 0.0182, -0.0081, 0.0197, ..., 0.0150, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0044, -0.0028, -0.0005, ..., 0.0003, 0.0015, -0.0123], + [-0.0150, 0.0251, 0.0432, ..., -0.0049, 0.0038, -0.0052], + [-0.0153, -0.0203, -0.0245, ..., -0.0120, -0.0325, 0.0040], + ..., + [ 0.0026, 0.0023, 0.0417, ..., 0.0280, 0.0030, -0.0042], + [-0.0024, 0.0001, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0065, -0.0019, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0155, -0.0062, -0.0481, ..., 0.0449, 0.0212, 0.0129], + [ 0.0154, -0.0227, 0.0220, ..., -0.0221, 0.0036, 0.0080], + [-0.0151, -0.0099, 0.0430, ..., -0.0095, 0.0327, -0.0111], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0139], + [-0.0017, 0.0166, -0.0059, ..., 0.0027, 0.0080, -0.0071], + [-0.0110, -0.0028, 0.0117, ..., 0.0058, -0.0034, -0.0075]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0040, -0.0386, -0.0118, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0079, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0126, -0.0044, -0.0126, ..., 0.0216, 0.0103, -0.0055], + ..., + [-0.0152, -0.0172, -0.0137, ..., 0.0098, 0.0098, 0.0184], + [-0.0152, -0.0356, 0.0201, ..., -0.0055, -0.0115, 0.0332], + [-0.0396, -0.0123, -0.0090, ..., -0.0284, -0.0142, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0277, 0.0312, 0.0133, ..., -0.0195, 0.0116, -0.0146], + [ 0.0192, 0.0095, -0.0075, ..., 0.0027, -0.0129, -0.0198], + [-0.0251, 0.0114, -0.0137, ..., 0.0093, 0.0172, 0.0075], + ..., + [ 0.0466, 0.0037, -0.0007, ..., -0.0044, 0.0239, 0.0031], + [ 0.0547, -0.0194, -0.0108, ..., -0.0056, 0.0311, -0.0023], + [ 0.0048, 0.0099, -0.0096, ..., 0.0203, 0.0170, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0045, -0.0032, 0.0283, ..., -0.0110, 0.0031, -0.0182], + [ 0.0049, 0.0129, -0.0315, ..., -0.0457, -0.0063, 0.0106], + [ 0.0027, 0.0261, 0.0245, ..., 0.0046, 0.0080, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0154, 0.0045, -0.0075], + [ 0.0055, -0.0371, 0.0356, ..., -0.0081, 0.0410, -0.0178], + [ 0.0178, -0.0151, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0079, 0.0208, 0.0273, ..., 0.0136, 0.0156, 0.0073], + [-0.0074, 0.0023, 0.0038, ..., 0.0146, 0.0015, 0.0050], + [ 0.0045, -0.0005, 0.0244, ..., -0.0120, -0.0055, -0.0247], + ..., + [ 0.0245, 0.0211, -0.0069, ..., 0.0247, -0.0092, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., -0.0052, -0.0295, 0.0400], + [-0.0116, 0.0132, 0.0305, ..., -0.0108, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0072, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0150, -0.0043, 0.0288, ..., 0.0141, 0.0583, -0.0025], + [-0.0281, 0.0156, -0.0386, ..., -0.0444, -0.0093, -0.0038], + ..., + [ 0.0041, 0.0157, 0.0217, ..., 0.0019, 0.0114, 0.0153], + [ 0.0059, 0.0195, -0.0283, ..., 0.0031, -0.0010, -0.0342], + [-0.0170, -0.0195, -0.0026, ..., 0.0049, 0.0165, 0.0133]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0139, -0.0146, 0.0459, ..., 0.0044, -0.0320, 0.0277], + [ 0.0033, 0.0064, 0.0270, ..., -0.0221, -0.0183, -0.0271], + [ 0.0221, 0.0219, -0.0201, ..., 0.0043, 0.0073, -0.0046], + ..., + [-0.0258, -0.0023, 0.0215, ..., 0.0041, 0.0237, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, 0.0014], + [ 0.0042, -0.0081, -0.0129, ..., -0.0334, 0.0060, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0145, -0.0454, ..., -0.0129, -0.0128, 0.0400], + [ 0.0295, -0.0187, -0.0109, ..., -0.0283, 0.0041, -0.0186], + [ 0.0052, 0.0547, 0.0033, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0280, -0.0058, -0.0269, ..., -0.0087, -0.0287, 0.0148], + [ 0.0091, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0098, 0.0104, ..., -0.0072, 0.0703, -0.0043]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0090, -0.0315, ..., 0.0078, 0.0254, 0.0359], + [ 0.0092, -0.0015, 0.0004, ..., -0.0123, -0.0233, -0.0046], + [-0.0025, 0.0046, -0.0126, ..., -0.0124, -0.0062, 0.0083], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0023], + [ 0.0339, -0.0072, 0.0223, ..., 0.0410, 0.0004, 0.0136], + [-0.0034, -0.0439, -0.0486, ..., 0.0139, -0.0129, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[-0.0035, 0.0141, -0.0068, ..., -0.0052, -0.0043, -0.0229], + [-0.0287, 0.0231, 0.0079, ..., -0.0142, -0.0036, 0.0222], + [ 0.0071, 0.0297, -0.0189, ..., -0.0168, 0.0311, -0.0262], + ..., + [ 0.0233, -0.0049, -0.0050, ..., 0.0045, 0.0278, -0.0034], + [-0.0103, 0.0035, 0.0142, ..., -0.0175, 0.0444, 0.0425], + [ 0.0078, -0.0079, 0.0099, ..., 0.0061, -0.0344, 0.0229]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[-0.0027, 0.0123, 0.0190, ..., -0.0096, -0.0198, 0.0148], + [-0.0581, 0.0294, 0.0032, ..., -0.0240, 0.0071, 0.0302], + [ 0.0087, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0040, -0.0289, 0.0537, ..., 0.0051, -0.0061, -0.0098], + [-0.0041, -0.0081, 0.0209, ..., 0.0049, 0.0101, 0.0186], + [-0.0087, -0.0349, -0.0173, ..., -0.0400, 0.0037, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0181, -0.0265, -0.0108, ..., 0.0096, -0.0133, -0.0405], + [ 0.0078, -0.0090, 0.0266, ..., -0.0400, 0.0073, 0.0391], + [-0.0045, 0.0359, 0.0284, ..., 0.0297, -0.0232, 0.0212], + ..., + [-0.0045, 0.0124, 0.0132, ..., -0.0134, 0.0016, 0.0095], + [ 0.0298, -0.0167, -0.0038, ..., -0.0070, 0.0048, -0.0206], + [ 0.0050, -0.0056, -0.0040, ..., 0.0009, -0.0044, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, 0.0002, 0.0123, ..., 0.0014, -0.0208, -0.0079], + [-0.0096, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0229], + [ 0.0137, -0.0278, 0.0117, ..., 0.0396, -0.0085, -0.0078], + ..., + [-0.0033, 0.0037, -0.0311, ..., 0.0344, 0.0062, -0.0124], + [-0.0236, -0.0100, -0.0067, ..., 0.0332, -0.0117, -0.0173], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0138, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[ 0.0038, 0.0166, 0.0041, ..., 0.0273, -0.0352, 0.0149], + [-0.0510, -0.0076, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0076, ..., -0.0388, 0.0004, 0.0398], + ..., + [-0.0160, -0.0195, -0.0052, ..., -0.0007, -0.0054, -0.0038], + [-0.0187, 0.0040, 0.0094, ..., 0.0287, -0.0122, -0.0098], + [-0.0181, -0.0038, 0.0247, ..., -0.0146, 0.0068, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0059, 0.0097, 0.0140, ..., -0.0112, -0.0019, -0.0253], + [-0.0166, 0.0330, -0.0171, ..., -0.0034, -0.0403, -0.0110], + [ 0.0084, 0.0347, 0.0275, ..., 0.0150, 0.0067, 0.0080], + ..., + [ 0.0271, -0.0167, -0.0106, ..., -0.0098, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0052, ..., 0.0033, 0.0017, -0.0243], + [-0.0079, 0.0581, 0.0067, ..., 0.0147, 0.0239, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0123, -0.0410, 0.0126, ..., -0.0254, -0.0027, -0.0173], + [ 0.0476, 0.0315, 0.0150, ..., -0.0146, -0.0070, 0.0038], + [-0.0092, -0.0270, 0.0086, ..., 0.0332, 0.0148, -0.0075], + ..., + [-0.0067, 0.0028, 0.0162, ..., -0.0177, 0.0171, -0.0134], + [ 0.0048, -0.0187, -0.0052, ..., 0.0248, -0.0140, -0.0058], + [-0.0142, 0.0022, 0.0150, ..., 0.0029, -0.0024, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0063, 0.0093, ..., 0.0255, -0.0253, -0.0132], + [-0.0048, -0.0135, -0.0160, ..., -0.0107, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0172, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0119, 0.0037, 0.0322, ..., -0.0183, -0.0033, 0.0031], + [ 0.0043, 0.0008, 0.0162, ..., 0.0028, 0.0255, 0.0071], + [ 0.0270, 0.0085, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0066, ..., -0.0532, 0.0104, -0.0042], + [-0.0240, -0.0215, -0.0103, ..., -0.0060, 0.0098, -0.0168], + [ 0.0062, -0.0046, -0.0064, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0059, 0.0172, -0.0106, ..., -0.0300, -0.0027, 0.0435], + [ 0.0049, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0066], + [-0.0167, 0.0123, 0.0101, ..., -0.0247, 0.0119, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0072, -0.0079, 0.0317, ..., 0.0199, -0.0105, -0.0376], + [ 0.0356, 0.0378, 0.0134, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., -0.0002, 0.0059, -0.0322], + ..., + [ 0.0059, 0.0107, -0.0040, ..., 0.0366, -0.0206, 0.0374], + [ 0.0024, -0.0079, 0.0040, ..., -0.0099, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0039, ..., -0.0166, -0.0059, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0048, -0.0087, -0.0344, ..., 0.0047, 0.0347, -0.0178], + [ 0.0267, -0.0366, -0.0089, ..., 0.0217, 0.0349, -0.0028], + [-0.0094, -0.0286, 0.0042, ..., -0.0026, -0.0282, 0.0029], + ..., + [-0.0128, 0.0037, 0.0053, ..., -0.0084, 0.0165, 0.0075], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0143], + [-0.0139, 0.0034, -0.0170, ..., -0.0128, 0.0061, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0028, 0.0034, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0118, 0.0374, ..., -0.0153, -0.0188, -0.0136], + ..., + [ 0.0262, -0.0019, -0.0075, ..., -0.0011, 0.0038, -0.0165], + [-0.0222, -0.0154, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0132, -0.0148, 0.0029, ..., 0.0020, -0.0234, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0102], + [ 0.0249, 0.0205, 0.0072, ..., -0.0032, -0.0092, -0.0103], + [-0.0220, -0.0022, -0.0023, ..., 0.0306, 0.0381, -0.0015], + ..., + [ 0.0200, -0.0011, 0.0170, ..., 0.0115, 0.0320, -0.0157], + [ 0.0349, 0.0117, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0067, 0.0256, -0.0023, ..., 0.0205, -0.0045, -0.0106]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0039, 0.0199, ..., 0.0013, 0.0276, -0.0312], + [ 0.0178, 0.0033, 0.0099, ..., 0.0110, -0.0398, 0.0293], + [ 0.0430, -0.0128, 0.0062, ..., 0.0040, -0.0084, 0.0067], + ..., + [ 0.0192, -0.0111, -0.0108, ..., -0.0045, 0.0330, 0.0140], + [-0.0126, -0.0542, -0.0205, ..., 0.0017, -0.0159, 0.0320], + [ 0.0303, 0.0066, -0.0425, ..., -0.0037, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0065, -0.0315, ..., 0.0525, 0.0039, -0.0349], + [-0.0019, -0.0039, -0.0154, ..., 0.0053, -0.0240, 0.0305], + [-0.0256, -0.0028, 0.0072, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0069, 0.0231, 0.0209, ..., 0.0075, 0.0045, 0.0461], + [-0.0072, -0.0079, -0.0001, ..., -0.0217, 0.0347, 0.0039], + [ 0.0032, 0.0234, 0.0038, ..., 0.0261, -0.0033, 0.0100]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0045, 0.0035], + [-0.0183, -0.0211, -0.0037, ..., -0.0145, 0.0138, -0.0162], + [-0.0066, -0.0160, 0.0070, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0096, 0.0078, ..., 0.0171, 0.0040, -0.0053], + [-0.0141, 0.0260, -0.0084, ..., 0.0078, 0.0103, -0.0166], + [ 0.0093, -0.0142, -0.0018, ..., 0.0053, 0.0056, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0070, 0.0141, -0.0053, ..., -0.0278, 0.0240, -0.0115], + [-0.0036, 0.0188, -0.0160, ..., -0.0317, -0.0043, 0.0181], + [-0.0006, 0.0310, -0.0085, ..., -0.0226, 0.0038, 0.0334], + ..., + [-0.0079, 0.0186, -0.0056, ..., 0.0228, 0.0052, 0.0046], + [ 0.0271, 0.0119, 0.0354, ..., 0.0019, 0.0198, 0.0143], + [ 0.0282, 0.0171, 0.0317, ..., -0.0110, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0114, -0.0265, -0.0110, ..., 0.0049, 0.0037, 0.0134], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0081], + [-0.0076, -0.0190, -0.0034, ..., 0.0166, -0.0135, 0.0137], + ..., + [-0.0200, 0.0018, 0.0060, ..., -0.0339, 0.0361, -0.0032], + [-0.0046, -0.0250, 0.0154, ..., -0.0022, -0.0162, 0.0220], + [-0.0044, 0.0038, 0.0205, ..., -0.0064, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0020, 0.0240, -0.0217, ..., 0.0172, -0.0028, 0.0115], + [-0.0386, 0.0103, 0.0074, ..., -0.0041, 0.0322, 0.0087], + [-0.0427, 0.0025, -0.0056, ..., 0.0068, 0.0110, -0.0139], + ..., + [-0.0084, -0.0043, -0.0264, ..., -0.0586, -0.0630, -0.0131], + [ 0.0064, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0106, -0.0610, 0.0162, ..., -0.0291, -0.0166, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0118, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0112], + [ 0.0108, -0.0101, -0.0327, ..., 0.0447, 0.0053, 0.0043], + [-0.0023, -0.0175, 0.0070, ..., -0.0076, -0.0049, 0.0026], + ..., + [ 0.0295, -0.0049, -0.0106, ..., 0.0095, -0.0160, -0.0288], + [ 0.0082, 0.0253, -0.0177, ..., 0.0095, 0.0071, 0.0066], + [-0.0135, -0.0057, 0.0137, ..., -0.0128, 0.0277, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0132, 0.0053, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0016, 0.0234, 0.0062, ..., 0.0464, 0.0069, 0.0309], + ..., + [-0.0118, -0.0175, 0.0327, ..., 0.0132, 0.0014, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, -0.0040, -0.0444, ..., 0.0060, -0.0283, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0081, -0.0101, 0.0146, ..., -0.0742, -0.0077, 0.0161], + [-0.0066, -0.0132, -0.0062, ..., 0.0039, -0.0023, 0.0238], + [ 0.0189, 0.0221, -0.0088, ..., 0.0018, 0.0408, 0.0045], + ..., + [ 0.0046, 0.0247, 0.0396, ..., 0.0043, -0.0304, -0.0183], + [ 0.0299, -0.0112, -0.0417, ..., -0.0039, -0.0376, -0.0045], + [-0.0210, 0.0168, -0.0050, ..., 0.0045, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0044, 0.0242, -0.0042, ..., -0.0245, 0.0033, -0.0070], + [ 0.0190, -0.0121, 0.0332, ..., -0.0144, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0037, ..., -0.0030, -0.0091, 0.0182], + ..., + [-0.0430, -0.0403, -0.0083, ..., 0.0132, -0.0141, 0.0055], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0090, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0050, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0247, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0093, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0061, ..., 0.0101, -0.0223, -0.0012], + ..., + [ 0.0192, 0.0292, -0.0018, ..., 0.0176, -0.0197, 0.0019], + [ 0.0265, 0.0063, 0.0056, ..., 0.0286, 0.0021, -0.0267], + [-0.0184, -0.0088, -0.0045, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0084, -0.0159, 0.0082, ..., 0.0198, 0.0051, 0.0277], + [ 0.0172, 0.0041, -0.0041, ..., -0.0003, 0.0040, 0.0067], + [ 0.0040, 0.0184, 0.0114, ..., 0.0099, -0.0053, -0.0059], + ..., + [ 0.0042, -0.0111, 0.0045, ..., -0.0096, 0.0036, 0.0121], + [-0.0427, -0.0391, -0.0192, ..., 0.0092, -0.0238, -0.0042], + [ 0.0038, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0083, -0.0161, 0.0123], + [ 0.0021, -0.0044, 0.0312, ..., 0.0071, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0071, 0.0046, 0.0278], + ..., + [ 0.0170, -0.0047, 0.0471, ..., -0.0057, 0.0038, -0.0079], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0039, -0.0203], + [ 0.0057, 0.0064, 0.0054, ..., -0.0095, 0.0013, 0.0021]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[ 0.0057, -0.0155, 0.0284, ..., -0.0104, 0.0135, -0.0052], + [ 0.0153, 0.0332, 0.0337, ..., 0.0223, -0.0165, 0.0038], + [-0.0020, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0069, 0.0062, 0.0280, ..., -0.0063, -0.0039, 0.0140], + [-0.0048, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0091, ..., -0.0035, 0.0167, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0045, ..., -0.0238, -0.0102, 0.0119], + [-0.0250, -0.0162, -0.0023, ..., -0.0084, 0.0002, -0.0369], + [ 0.0010, 0.0141, 0.0075, ..., -0.0452, -0.0045, 0.0120], + ..., + [ 0.0147, -0.0141, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0099, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0142, 0.0034, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[-0.0069, -0.0327, -0.0190, ..., 0.0042, 0.0036, -0.0209], + [ 0.0713, 0.0151, -0.0033, ..., 0.0051, 0.0186, 0.0020], + [-0.0227, -0.0092, 0.0188, ..., -0.0122, 0.0193, 0.0075], + ..., + [ 0.0116, 0.0464, 0.0083, ..., 0.0248, 0.0168, -0.0039], + [ 0.0381, -0.0036, 0.0060, ..., 0.0046, -0.0109, 0.0061], + [ 0.0292, 0.0050, 0.0209, ..., -0.0036, 0.0234, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0042, 0.0212, ..., 0.0305, 0.0137, 0.0104], + [-0.0060, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0065], + ..., + [ 0.0084, -0.0275, 0.0225, ..., -0.0039, -0.0189, -0.0107], + [-0.0125, 0.0041, -0.0045, ..., 0.0045, 0.0040, 0.0211], + [ 0.0020, -0.0060, -0.0059, ..., -0.0109, 0.0264, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[ 0.0043, -0.0076, -0.0108, ..., 0.0186, -0.0216, 0.0049], + [ 0.0039, 0.0123, -0.0231, ..., 0.0123, -0.0112, -0.0183], + [ 0.0107, 0.0103, -0.0011, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0123, 0.0146, ..., 0.0540, 0.0415, 0.0097], + [ 0.0038, 0.0087, -0.0138, ..., 0.0055, -0.0189, -0.0237], + [-0.0019, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, 0.0066, 0.0359, ..., -0.0084, -0.0121, -0.0041], + [ 0.0190, -0.0044, 0.0298, ..., 0.0126, 0.0048, -0.0200], + [ 0.0118, -0.0110, 0.0177, ..., -0.0248, 0.0229, -0.0038], + ..., + [-0.0077, -0.0048, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0025, -0.0291, ..., 0.0206, 0.0304, -0.0085], + [-0.0077, 0.0245, -0.0008, ..., -0.0413, 0.0216, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0060, -0.0157, -0.0032, ..., 0.0047, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0072], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, 0.0049], + ..., + [ 0.0033, -0.0058, 0.0258, ..., -0.0310, 0.0092, -0.0361], + [ 0.0040, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [-0.0036, 0.0157, -0.0046, ..., 0.0052, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0070, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0105, ..., 0.0104, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0037, ..., 0.0129, -0.0047, 0.0087], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [ 0.0038, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0128, 0.0206, ..., 0.0058, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0094, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0060, 0.0066, ..., 0.0015, 0.0112, 0.0598], + [-0.0060, 0.0267, -0.0065, ..., 0.0284, -0.0055, -0.0364], + ..., + [-0.0052, 0.0002, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, -0.0036, -0.0089, ..., 0.0211, 0.0134, -0.0078], + [ 0.0102, 0.0236, -0.0206, ..., -0.0134, 0.0048, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0077, 0.0294, ..., 0.0031, 0.0270, 0.0063], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0066, -0.0168], + [ 0.0201, -0.0121, 0.0126, ..., -0.0381, -0.0247, -0.0070], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0078], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0068, -0.0466], + [ 0.0327, 0.0284, -0.0039, ..., 0.0145, 0.0139, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, 0.0039, -0.0167, ..., 0.0005, 0.0170, 0.0107], + [ 0.0038, 0.0009, 0.0295, ..., -0.0172, 0.0038, 0.0135], + [ 0.0193, 0.0236, -0.0125, ..., -0.0045, -0.0369, 0.0239], + ..., + [ 0.0154, -0.0413, -0.0123, ..., -0.0354, 0.0062, -0.0320], + [-0.0266, -0.0071, -0.0327, ..., 0.0040, 0.0049, 0.0043], + [ 0.0298, 0.0356, -0.0192, ..., 0.0055, 0.0039, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0045, 0.0043, 0.0317, ..., 0.0053, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0071, -0.0096], + [-0.0166, 0.0059, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, 0.0049, 0.0260, ..., 0.0159, 0.0054, -0.0474], + [-0.0042, 0.0080, 0.0038, ..., 0.0143, 0.0557, 0.0023], + [-0.0203, 0.0273, 0.0029, ..., -0.0330, 0.0049, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0074, ..., 0.0036, -0.0093, -0.0277], + [ 0.0049, -0.0041, -0.0038, ..., -0.0049, 0.0140, -0.0161], + [-0.0049, 0.0168, -0.0057, ..., -0.0092, 0.0124, 0.0039], + ..., + [ 0.0398, 0.0452, 0.0251, ..., -0.0076, -0.0330, -0.0043], + [-0.0620, 0.0033, 0.0371, ..., 0.0410, 0.0037, -0.0173], + [ 0.0126, -0.0039, -0.0270, ..., -0.0037, 0.0388, -0.0104]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, -0.0039, 0.0245], + [-0.0449, 0.0039, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0061, ..., 0.0029, 0.0288, 0.0124], + [ 0.0330, 0.0040, 0.0059, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0074, -0.0030, ..., -0.0063, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0106, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0045, 0.0029, 0.0527], + [ 0.0041, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0086], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0095, -0.0095, 0.0327], + [ 0.0142, 0.0114, 0.0061, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0126, 0.0281, ..., -0.0242, -0.0126, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0029], + [-0.0053, 0.0039, -0.0286, ..., 0.0054, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0057, ..., -0.0037, 0.0081, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0096, 0.0150], + [-0.0025, 0.0021, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0071, ..., 0.0092, -0.0114, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0070, -0.0017, ..., -0.0220, -0.0145, 0.0040], + [-0.0166, 0.0189, -0.0065, ..., 0.0096, -0.0018, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0006, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0071, ..., -0.0041, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0070, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0122, 0.0289, 0.0125, ..., 0.0134, 0.0078, -0.0103], + [ 0.0188, -0.0197, -0.0039, ..., 0.0284, -0.0342, -0.0072], + [-0.0525, -0.0215, 0.0154, ..., -0.0270, 0.0084, 0.0254], + ..., + [-0.0068, 0.0087, 0.0281, ..., 0.0129, -0.0043, -0.0045], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0038, 0.0026], + [-0.0273, -0.0747, 0.0208, ..., 0.0041, 0.0041, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0047, 0.0150, 0.0334, ..., 0.0205, -0.0273, 0.0040], + [-0.0165, 0.0111, -0.0135, ..., 0.0106, -0.0069, 0.0219], + [ 0.0053, 0.0227, -0.0056, ..., 0.0128, -0.0070, 0.0150], + ..., + [-0.0347, -0.0148, 0.0061, ..., 0.0021, -0.0021, -0.0093], + [-0.0065, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0024, -0.0500, -0.0232, ..., 0.0320, 0.0022, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0059, -0.0364], + [ 0.0220, 0.0036, -0.0114, ..., 0.0047, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, -0.0039, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0074, 0.0192, -0.0393], + [-0.0101, -0.0244, 0.0275, ..., -0.0057, 0.0044, 0.0081], + [-0.0170, -0.0427, 0.0140, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0046, -0.0378, -0.0359, ..., 0.0552, -0.0073, -0.0160], + [ 0.0017, 0.0142, 0.0134, ..., 0.0049, -0.0153, 0.0354], + [ 0.0095, -0.0038, 0.0057, ..., -0.0016, -0.0249, -0.0135], + ..., + [ 0.0078, -0.0126, -0.0059, ..., -0.0095, -0.0084, -0.0028], + [-0.0059, 0.0439, 0.0166, ..., 0.0248, 0.0132, -0.0187], + [ 0.0175, 0.0060, 0.0043, ..., 0.0410, -0.0096, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0107, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0143, -0.0031, ..., -0.0039, -0.0244, 0.0356], + [ 0.0077, 0.0165, -0.0200, ..., -0.0048, 0.0237, -0.0071], + ..., + [-0.0197, -0.0022, 0.0386, ..., 0.0181, 0.0112, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, 0.0035, -0.0374], + [-0.0084, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0047, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0136, 0.0114, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, -0.0032, -0.0040, ..., -0.0048, -0.0030, 0.0337], + ..., + [ 0.0422, 0.0046, 0.0352, ..., -0.0304, -0.0039, -0.0204], + [ 0.0302, 0.0156, 0.0103, ..., -0.0076, -0.0254, -0.0245], + [-0.0041, -0.0069, 0.0042, ..., -0.0045, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, 0.0040, 0.0266, ..., 0.0154, 0.0187, 0.0039], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0041, ..., -0.0172, -0.0091, 0.0231], + ..., + [-0.0295, -0.0137, 0.0001, ..., -0.0041, -0.0052, 0.0099], + [-0.0045, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0075], + [ 0.0040, -0.0048, -0.0259, ..., 0.0045, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[-0.0047, 0.0062, -0.0140, ..., 0.0242, -0.0040, -0.0386], + [-0.0141, 0.0064, -0.0097, ..., -0.0369, -0.0155, 0.0157], + [-0.0267, -0.0020, -0.0068, ..., -0.0127, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0019, 0.0144, ..., 0.0115, 0.0039, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., 0.0041, -0.0039, -0.0137], + [-0.0219, -0.0396, -0.0078, ..., 0.0028, 0.0041, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0036, 0.0168, 0.0194, ..., 0.0349, -0.0420, 0.0013], + [ 0.0043, 0.0273, 0.0146, ..., -0.0184, -0.0035, 0.0154], + [ 0.0145, -0.0039, -0.0078, ..., 0.0245, 0.0359, 0.0059], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0037, 0.0044, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0073, -0.0312, -0.0457], + [ 0.0056, 0.0034, -0.0057, ..., 0.0021, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0037, 0.0121, 0.0039, ..., 0.0126, 0.0018, 0.0074], + [-0.0277, -0.0030, -0.0181, ..., 0.0038, -0.0029, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0043, 0.0513, -0.0113], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0113], + [-0.0282, -0.0042, 0.0121, ..., -0.0045, -0.0017, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0032, -0.0130, ..., 0.0141, 0.0308, 0.0056], + [ 0.0120, -0.0332, -0.0320, ..., -0.0073, -0.0142, -0.0032], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, 0.0045, -0.0311, ..., 0.0120, 0.0058, 0.0082], + [ 0.0043, 0.0198, 0.0141, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0125, -0.0297, -0.0055, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0103, ..., 0.0165, -0.0280, 0.0481], + [-0.0092, 0.0043, 0.0095, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, 0.0019, ..., 0.0302, -0.0025, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0039, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0022, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0048, -0.0222, 0.0262], + ..., + [-0.0147, 0.0049, 0.0260, ..., 0.0044, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0056, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0046, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0147, -0.0249, 0.0114], + [-0.0278, -0.0078, -0.0374, ..., 0.0276, -0.0062, 0.0150], + [ 0.0088, 0.0237, -0.0156, ..., 0.0527, -0.0075, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0072, -0.0095], + [-0.0071, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0034, -0.0129, ..., 0.0039, -0.0371, 0.0043], + [ 0.0059, -0.0127, 0.0043, ..., -0.0204, 0.0039, 0.0065], + [-0.0112, -0.0195, 0.0052, ..., -0.0542, -0.0089, -0.0045], + ..., + [-0.0205, -0.0084, 0.0038, ..., -0.0199, 0.0049, 0.0036], + [-0.0220, -0.0302, -0.0103, ..., 0.0107, 0.0369, -0.0038], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0039, 0.0049, 0.0238, ..., -0.0146, -0.0320, -0.0021], + [-0.0193, 0.0364, -0.0029, ..., -0.0061, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0063, -0.0232, -0.0464], + [-0.0110, -0.0298, -0.0195, ..., 0.0112, 0.0312, 0.0152], + [-0.0070, -0.0119, -0.0195, ..., -0.0042, 0.0109, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0060, -0.0239, -0.0123, ..., -0.0040, 0.0243, -0.0396], + [-0.0100, 0.0153, 0.0061, ..., -0.0277, -0.0045, 0.0016], + ..., + [ 0.0118, 0.0053, 0.0515, ..., 0.0327, -0.0118, -0.0064], + [-0.0040, 0.0170, -0.0006, ..., -0.0066, 0.0280, -0.0210], + [ 0.0012, 0.0075, 0.0041, ..., 0.0154, -0.0061, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0061, 0.0081, -0.0074, ..., -0.0588, 0.0187, -0.0067], + [-0.0103, 0.0199, -0.0078, ..., 0.0080, -0.0332, -0.0043], + ..., + [ 0.0045, -0.0464, 0.0277, ..., -0.0142, 0.0198, -0.0075], + [ 0.0162, 0.0075, 0.0089, ..., -0.0106, 0.0044, 0.0297], + [-0.0270, 0.0132, 0.0166, ..., 0.0081, 0.0175, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0053, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, -0.0007, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, -0.0039, -0.0201, ..., 0.0330, 0.0309, -0.0046], + ..., + [-0.0037, 0.0123, 0.0153, ..., -0.0042, -0.0037, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0039, 0.0175], + [-0.0244, -0.0058, -0.0300, ..., -0.0048, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0017, -0.0058, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, -0.0039, ..., 0.0310, 0.0088, -0.0037], + [ 0.0042, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0109, -0.0237, ..., -0.0104, 0.0157, 0.0064], + [ 0.0079, 0.0205, 0.0146, ..., 0.0262, -0.0195, 0.0082], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0080]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0029, ..., -0.0053, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., 0.0013, 0.0034, -0.0093], + [ 0.0017, 0.0002, 0.0398, ..., -0.0304, -0.0030, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0035, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0040], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0047, 0.0291], + [ 0.0132, -0.0035, -0.0339, ..., 0.0066, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, 0.0036], + [-0.0166, -0.0037, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0062, -0.0039, ..., 0.0240, -0.0037, -0.0031]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0007], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [ 0.0010, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0060, -0.0094], + [ 0.0059, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0078], + [-0.0206, 0.0417, -0.0125, ..., -0.0071, 0.0041, -0.0039], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0041, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, 0.0042, ..., -0.0164, 0.0036, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0028, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0054, -0.0231, -0.0255, ..., -0.0045, 0.0295, -0.0063], + [-0.0255, 0.0513, 0.0039, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0033, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0077], + [-0.0065, -0.0061, -0.0364, ..., -0.0175, 0.0306, -0.0077], + [-0.0051, -0.0195, 0.0019, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0075, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0067, -0.0193, -0.0140], + [-0.0166, -0.0048, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0128, -0.0076, 0.0140, ..., -0.0073, 0.0211, -0.0275], + [-0.0039, -0.0077, -0.0234, ..., 0.0087, -0.0070, -0.0166], + [ 0.0225, -0.0251, 0.0028, ..., 0.0078, 0.0236, 0.0081], + ..., + [ 0.0186, 0.0144, -0.0053, ..., 0.0272, -0.0084, -0.0092], + [-0.0040, -0.0030, 0.0071, ..., -0.0260, 0.0267, 0.0147], + [ 0.0101, 0.0009, 0.0212, ..., 0.0195, -0.0157, 0.0121]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0087, -0.0270, 0.0253, ..., 0.0099, 0.0048, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.1.proj.1.weight': tensor([[-0.0043, 0.0179, 0.0830, ..., -0.0645, 0.0269, -0.0361], + [-0.0288, -0.0048, -0.1138, ..., 0.0220, -0.0299, 0.0554], + [-0.0253, 0.1050, -0.0269, ..., -0.0189, 0.0175, -0.0332], + ..., + [ 0.0144, 0.0107, -0.0055, ..., 0.0226, -0.0234, 0.0046], + [-0.0284, -0.0649, 0.0503, ..., 0.0425, 0.0728, 0.0615], + [ 0.0162, -0.0171, 0.0017, ..., 0.0282, 0.0413, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_proj.weight': tensor([[-0.0054, -0.0018, 0.0101, ..., 0.0036, 0.0021, 0.0046], + [ 0.0051, 0.0034, -0.0007, ..., -0.0087, 0.0004, -0.0032], + [ 0.0102, -0.0041, 0.0022, ..., -0.0049, -0.0089, -0.0043], + ..., + [-0.0117, 0.0017, -0.0068, ..., 0.0032, 0.0063, -0.0047], + [ 0.0031, -0.0067, -0.0085, ..., -0.0039, -0.0055, -0.0053], + [ 0.0129, -0.0121, 0.0019, ..., 0.0051, -0.0067, 0.0019]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_norm.weight': tensor([ 1.5918e-01, -1.1572e-01, 1.1377e-01, 2.5977e-01, 7.2002e-05, + 5.5313e-04, 8.7172e-07, 7.4133e-07, 2.3041e-03, 4.2564e-10, + 1.2878e-09, -4.3165e-13, 1.2719e-12, 3.6948e-13, 4.2810e-13, + 7.7804e-13, 5.3478e-10, -1.0014e-05, -8.0327e-09, 3.6863e-17, + -1.7675e-13, 1.2612e-13, 2.4414e-02, 3.3379e-04, 2.8320e-02, + 6.4453e-01, 1.0400e-01, 4.0283e-03, 5.8838e-02, 9.5215e-02, + -1.3434e-14, -2.3761e-11, 6.0390e-10, -3.6380e-10, 8.9585e-11, + -6.4448e-07, 3.3833e-10, 2.2315e-14, 2.8848e-12, -1.6225e-09, + -5.2296e-12, 7.5033e-11, -5.2736e-15, 9.5367e-05, 2.4536e-02, + 6.1417e-04, -5.3955e-02, 5.8594e-02, 4.5117e-01, -1.0529e-03, + -1.8539e-03, 8.6799e-07, -3.5390e-08, -2.7008e-07, 2.2965e-11, + 9.8225e-10, -2.4556e-10, 1.8813e-07, 2.0227e-09, -2.2055e-11, + -2.7512e-18, -9.1073e-17, -3.8337e-16, -5.5422e-13, 2.7344e-01, + 2.4707e-01, 3.8086e-01, 5.0781e-02, -6.8545e-06, -3.0365e-03, + -6.7521e-09, 3.1710e-05, 2.0294e-03, -2.5437e-12, -1.0959e-10, + 3.0909e-13, 8.7766e-11, 1.1050e-10, 3.5749e-14, 1.7337e-12, + 1.9063e-09, 7.6834e-09, -3.8654e-11, -7.3830e-15, -3.1225e-15, + 1.7167e-11, 3.7537e-03, -1.1673e-03, 6.0156e-01, 7.9102e-02, + 5.3125e-01, 9.1553e-03, 9.8419e-04, 2.2888e-03, 1.9984e-14, + -1.2733e-09, -6.7812e-09, -9.9590e-11, -9.9681e-10, 5.9605e-06, + -3.2152e-13, 4.4409e-16, 2.5757e-14, 2.2847e-09, 2.1982e-14, + 7.9581e-11, -1.2506e-11, 1.7456e-02, -1.8234e-03, -1.6098e-03, + 4.7656e-01, 3.2227e-01, 7.0312e-02, -1.2665e-03, -1.2665e-03, + 1.6928e-05, -3.5288e-10, -7.1526e-07, -1.4097e-10, -6.1846e-10, + -5.4570e-11, 2.3865e-08, 3.2742e-11, -1.5250e-08, 5.2259e-17, + -1.6697e-17, 1.9451e-13, -1.4766e-14], dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.self_attn.k_proj.weight': tensor([[ 0.0060, 0.0034, 0.0004, ..., -0.0048, -0.0052, 0.0039], + [-0.0184, 0.0046, 0.0019, ..., 0.0067, -0.0027, 0.0020], + [ 0.0284, -0.0114, 0.0126, ..., -0.0076, 0.0044, 0.0008], + ..., + [-0.0018, 0.0026, -0.0046, ..., -0.0025, 0.0008, 0.0019], + [-0.0019, 0.0022, -0.0029, ..., -0.0026, 0.0020, 0.0021], + [-0.0021, 0.0010, -0.0025, ..., 0.0011, 0.0010, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.k_norm.weight': tensor([ 2.6758e-01, 1.9043e-01, 1.1426e-01, 2.2461e-01, 4.5538e-05, + -1.8234e-03, -8.0466e-07, -3.5018e-06, 2.4414e-03, 3.0195e-10, + 1.3461e-09, -5.6133e-13, -6.3949e-12, -1.7208e-14, 3.1619e-13, + -7.6739e-13, 6.0754e-10, 1.0133e-05, -6.8976e-09, -8.2399e-18, + 2.0517e-13, 1.1369e-13, 1.9165e-02, -6.6757e-04, 5.0293e-02, + 6.4453e-01, 1.2109e-01, -8.1177e-03, 3.9062e-03, 3.8605e-03, + -7.0083e-16, -2.1737e-10, -2.8667e-09, 3.5652e-10, 5.1841e-11, + 1.2368e-06, 3.3833e-10, -2.3093e-14, -2.9132e-12, 2.0227e-09, + -5.2864e-12, -9.0495e-11, 9.9365e-15, 1.2817e-02, 2.0264e-02, + -7.3433e-05, 1.2451e-01, 1.2061e-01, 4.7656e-01, -1.0376e-03, + 1.9684e-03, -6.7949e-06, 4.1677e-08, -5.0291e-07, 1.8616e-12, + 5.7116e-10, 1.3006e-10, -1.8626e-07, -2.0664e-09, -2.6375e-11, + -3.6863e-18, -2.9317e-16, -3.1398e-16, -5.5778e-13, 8.3496e-02, + 2.5391e-01, 3.8281e-01, -1.7383e-01, 3.1948e-05, -2.0905e-03, + -5.6578e-08, -2.1458e-06, -4.7302e-04, 1.9372e-10, 4.5929e-11, + -2.4514e-13, -7.7307e-11, 1.1914e-10, 6.5725e-14, 9.0949e-13, + -1.7026e-09, 1.0710e-08, -2.9786e-11, 2.7534e-14, 5.1903e-15, + 1.6485e-11, 2.5482e-03, -3.2043e-04, 5.9766e-01, -7.2266e-02, + 5.3125e-01, -4.3640e-03, 5.8105e-02, 9.4238e-02, -2.6201e-14, + 1.2005e-09, 5.5879e-09, 8.9130e-11, 1.0332e-09, -5.3644e-06, + -3.5882e-13, 1.4988e-15, -2.4691e-13, -2.3138e-09, -5.6399e-14, + -7.9581e-11, -1.2449e-11, 1.2390e-02, 1.6479e-02, 1.2360e-03, + 4.7656e-01, 2.9883e-01, 9.6680e-02, 9.9945e-04, -1.9264e-04, + 1.4544e-05, 3.4197e-09, 1.4668e-08, 1.3824e-10, -9.9681e-10, + -1.0823e-10, -2.7358e-08, -5.7128e-12, 1.5483e-08, 5.2692e-17, + -7.6978e-18, 1.9362e-13, -1.5721e-13], dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.self_attn.v_proj.weight': tensor([[ 0.0034, -0.0011, 0.0074, ..., -0.0056, 0.0005, -0.0013], + [ 0.0107, 0.0002, 0.0008, ..., 0.0023, -0.0015, -0.0035], + [ 0.0165, 0.0069, 0.0038, ..., -0.0015, 0.0008, 0.0028], + ..., + [ 0.0009, -0.0049, 0.0047, ..., 0.0081, -0.0025, -0.0019], + [-0.0013, -0.0104, -0.0045, ..., 0.0058, -0.0014, 0.0033], + [-0.0010, 0.0017, -0.0020, ..., -0.0033, 0.0019, -0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.output_proj.weight': tensor([[ 6.3171e-03, -2.3071e-02, -2.7222e-02, ..., -3.5858e-04, + -2.0599e-03, 8.5831e-04], + [-3.2043e-03, -8.5449e-03, 1.1353e-02, ..., 2.3499e-03, + 1.0925e-02, -5.2185e-03], + [ 4.8256e-04, 3.1128e-03, -2.5024e-03, ..., -8.8120e-04, + -5.2490e-03, 2.2888e-03], + ..., + [ 3.9673e-03, 5.1270e-03, 7.6294e-03, ..., 1.1444e-03, + -4.1809e-03, -7.2002e-05], + [-1.2085e-02, 7.8125e-03, 8.2397e-03, ..., -2.8687e-03, + 3.0212e-03, 1.2024e-02], + [-6.0425e-03, 4.6997e-03, -9.2773e-03, ..., -1.0071e-02, + 3.3528e-06, 5.1880e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_proj.weight': tensor([[-0.0041, -0.0047, -0.0041, ..., 0.0022, -0.0011, -0.0018], + [-0.0050, -0.0005, 0.0062, ..., 0.0010, -0.0060, -0.0040], + [ 0.0042, 0.0014, 0.0048, ..., -0.0018, 0.0037, 0.0048], + ..., + [-0.0006, 0.0045, -0.0013, ..., 0.0008, -0.0028, 0.0034], + [-0.0029, -0.0054, -0.0033, ..., -0.0016, -0.0026, -0.0128], + [ 0.0023, -0.0041, 0.0075, ..., -0.0037, 0.0009, 0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_norm.weight': tensor([ 1.7700e-03, 1.3447e-04, 1.0885e-08, 1.7319e-03, -2.4319e-05, + 2.3346e-03, -4.5896e-06, 4.3030e-03, 4.0283e-02, 2.7466e-02, + 3.2663e-05, 3.4424e-02, 5.6641e-01, 3.1586e-03, 2.4902e-02, + 1.1921e-04, 4.3750e-01, -1.3947e-05, 2.3041e-03, 5.0659e-03, + 1.1536e-02, 7.9727e-04, 9.2773e-02, 2.3041e-03, 1.0925e-02, + 4.1504e-03, 2.8687e-02, -1.3161e-04, 1.2268e-02, 1.9989e-03, + 3.6133e-01, 8.1787e-03, 2.6226e-05, 4.0820e-01, 5.3516e-01, + 3.3722e-03, 3.8910e-04, 3.8818e-02, 1.6895e-01, 3.1586e-03, + 3.3203e-02, 4.9133e-03, -4.6194e-07, 1.0010e-02, 4.0283e-03, + 3.8574e-02, 1.3794e-02, 1.0620e-02, 3.7689e-03, 9.6798e-05, + 9.5215e-03, 9.5963e-06, 1.9372e-06, 1.7262e-04, 2.2411e-04, + 2.5482e-03, 1.2207e-02, 8.6670e-03, 3.0518e-03, 8.3618e-03, + 5.8365e-04, -1.2779e-04, 1.0925e-02, 9.2163e-03, 2.0905e-03, + 1.2970e-03, 2.8992e-04, 2.6758e-01, 7.8125e-03, 1.6968e-02, + 3.9844e-01, 1.0586e-04, 4.3213e-02, 9.6130e-04, 3.6210e-06, + -1.0757e-03, 5.7068e-03, 8.3542e-04, 5.4932e-04, 1.3000e-02, + 3.7384e-03, 7.1106e-03, 1.8066e-01, 4.0894e-03, 1.0376e-03, + 2.4048e-02, 5.1270e-03, 4.1771e-04, 9.1171e-04, 4.7461e-01, + 4.7302e-03, 2.5330e-03, 1.4587e-02, -7.2956e-05, 2.7275e-04, + 4.2188e-01, 1.1396e-04, 2.2736e-03, -2.8558e-10, 1.1673e-03, + 4.0430e-01, 1.1683e-04, 9.8419e-04, 2.4223e-04, 7.5684e-03, + -1.0252e-04, 1.5106e-03, -2.0409e-04, 3.5248e-03, 5.7373e-03, + 1.1826e-03, 2.3365e-05, 8.4229e-03, 3.7231e-03, 5.0049e-03, + 5.3406e-04, 1.5564e-02, 1.8845e-03, 5.1270e-03, 5.2261e-04, + 7.3433e-05, 7.9956e-03, 5.4932e-03, 2.7847e-04, 1.3123e-03, + 3.4943e-03, 2.0409e-04, 3.8147e-04], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_proj.weight': tensor([[ 1.0452e-03, 7.2861e-04, 2.1057e-03, ..., -8.8120e-04, + -1.5717e-03, -8.1253e-04], + [-3.4790e-03, -2.0599e-03, -4.6997e-03, ..., -2.5787e-03, + -3.6621e-03, -2.4567e-03], + [-1.2512e-03, -1.4877e-03, -2.9907e-03, ..., -2.7466e-03, + -5.3406e-03, -2.0294e-03], + ..., + [ 2.0599e-03, -2.4719e-03, 2.5177e-03, ..., 1.4114e-03, + 1.0633e-04, 2.6703e-04], + [-8.0872e-04, 5.3024e-04, -2.4872e-03, ..., -3.3264e-03, + -5.9605e-05, 1.4877e-03], + [ 4.3106e-04, 1.8997e-03, -2.5024e-03, ..., -4.2114e-03, + 4.4632e-04, 1.0452e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_norm.weight': tensor([ 1.7624e-03, 1.3351e-04, -8.2946e-10, 1.4343e-03, 2.6703e-05, + 2.3499e-03, 5.7817e-06, 4.3030e-03, 3.9307e-02, 2.7466e-02, + 3.1710e-05, 3.4668e-02, 5.7031e-01, 3.1433e-03, 2.4414e-02, + 6.9141e-05, 4.3750e-01, -1.3828e-05, 2.1210e-03, 5.0049e-03, + 1.1536e-02, 8.9264e-04, 9.2773e-02, 2.4414e-03, 1.0864e-02, + 4.0894e-03, 2.8564e-02, -1.3065e-04, 1.1597e-02, 1.9989e-03, + 3.6133e-01, 7.7820e-03, 2.6464e-05, 4.0820e-01, 5.3516e-01, + 3.2654e-03, 5.8365e-04, 3.9062e-02, 1.6992e-01, 3.2959e-03, + 3.3203e-02, 4.8828e-03, -3.3900e-07, 9.8877e-03, 3.4943e-03, + 3.8574e-02, 1.3733e-02, 1.0559e-02, 3.7689e-03, 9.6321e-05, + 9.3994e-03, 9.8348e-06, 3.9339e-06, 1.8501e-04, 1.5869e-03, + 2.5787e-03, 1.2146e-02, 8.6670e-03, 3.5095e-03, 8.3618e-03, + 3.3379e-04, -1.2779e-04, 1.0925e-02, 9.0332e-03, 2.0905e-03, + 1.3351e-03, 2.8992e-04, 2.6758e-01, 7.7820e-03, 1.6968e-02, + 3.9844e-01, 1.0538e-04, 4.1260e-02, 8.8501e-04, 3.7402e-06, + 8.8501e-04, 4.8828e-03, 7.8964e-04, 5.4932e-04, 1.2878e-02, + 3.7689e-03, 7.0801e-03, 1.8066e-01, 3.8910e-03, 9.9945e-04, + 2.3926e-02, 6.3477e-03, 4.1771e-04, 8.8120e-04, 4.7461e-01, + 4.7607e-03, 2.4872e-03, 1.4282e-02, -7.2479e-05, 2.7275e-04, + 4.2188e-01, 6.0797e-05, 2.2583e-03, -3.5288e-10, 1.1597e-03, + 4.0430e-01, 1.2207e-04, 2.1362e-03, 2.3365e-04, 7.5378e-03, + -1.0347e-04, 1.5182e-03, -2.0409e-04, 2.8381e-03, 5.7373e-03, + 1.4877e-03, 3.7670e-05, 8.4229e-03, 3.6926e-03, 5.2185e-03, + -4.1199e-04, 1.5625e-02, 2.0142e-03, 6.0730e-03, 5.5313e-04, + -1.7548e-04, 8.0566e-03, 5.4321e-03, 1.6117e-04, 4.1425e-06, + 3.5095e-03, 2.9945e-04, 2.4319e-04], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.v_proj.weight': tensor([[-1.2390e-02, 3.1586e-03, 1.1902e-02, ..., 3.7842e-03, + 4.3640e-03, -1.0925e-02], + [-2.4658e-02, 2.7120e-06, 8.6060e-03, ..., 1.5030e-03, + -5.3101e-03, 6.1989e-05], + [-1.3000e-02, 3.0670e-03, 5.8594e-03, ..., 2.9602e-03, + 1.4648e-02, -1.2146e-02], + ..., + [ 9.6436e-03, 3.1090e-04, 1.8234e-03, ..., 1.4572e-03, + 5.8899e-03, 1.9043e-02], + [-9.1553e-03, 6.5613e-04, 1.0437e-02, ..., 1.2817e-03, + -1.7090e-02, -3.9368e-03], + [-1.5381e-02, -3.2654e-03, 5.3406e-03, ..., 1.8921e-03, + -8.4229e-03, -2.1362e-02]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.output_proj.weight': tensor([[ 0.0031, -0.0066, -0.0034, ..., -0.0017, 0.0056, -0.0074], + [ 0.0012, 0.0079, -0.0082, ..., 0.0134, 0.0061, -0.0087], + [-0.0018, -0.0019, -0.0030, ..., 0.0065, -0.0018, -0.0021], + ..., + [ 0.0019, 0.0023, 0.0046, ..., 0.0103, 0.0028, -0.0043], + [ 0.0082, -0.0067, 0.0049, ..., -0.0066, -0.0098, 0.0054], + [ 0.0154, -0.0197, -0.0110, ..., 0.0051, -0.0073, 0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_img.weight': tensor([[ 0.0057, 0.0117, -0.0008, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0077, 0.0249, 0.0291], + [-0.0092, 0.0061, -0.0121, ..., -0.0342, -0.0327, 0.0064], + ..., + [ 0.0070, -0.0040, 0.0016, ..., 0.0085, 0.0286, -0.0165], + [ 0.0071, -0.0060, -0.0100, ..., -0.0098, 0.0143, -0.0309], + [ 0.0242, -0.0104, 0.0255, ..., -0.0228, 0.0039, -0.0156]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0304, ..., -0.0220, 0.0247, 0.0033], + [ 0.0030, 0.0121, 0.0405, ..., 0.0128, -0.0410, 0.0361], + [-0.0089, 0.0113, -0.0087, ..., -0.0057, -0.0038, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0034, 0.0002], + [-0.0160, 0.0047, 0.0013, ..., 0.0134, -0.0175, -0.0044], + [-0.0026, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0102]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0048, -0.0364, ..., -0.0225, -0.0142, -0.0046], + [-0.0157, 0.0058, -0.0364, ..., 0.0099, 0.0119, 0.0079], + [-0.0040, 0.0615, 0.0044, ..., -0.0205, -0.0306, -0.0114], + ..., + [ 0.0038, -0.0040, -0.0103, ..., -0.0211, 0.0014, 0.0288], + [ 0.0266, -0.0117, -0.0052, ..., -0.0420, -0.0282, 0.0048], + [ 0.0286, 0.0325, 0.0097, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.mlp.layer1.weight': tensor([[ 5.4932e-03, -5.7373e-03, 4.5776e-03, ..., -4.4250e-03, + 3.2349e-03, -4.3945e-03], + [ 4.7607e-03, -3.5858e-03, -1.3351e-03, ..., 2.7008e-03, + -3.3722e-03, -3.7079e-03], + [ 3.9062e-03, 6.5308e-03, -1.2684e-04, ..., 9.3460e-04, + 2.3041e-03, -3.6469e-03], + ..., + [-5.8289e-03, -6.7444e-03, 3.8719e-04, ..., -9.0332e-03, + 6.6528e-03, -6.7749e-03], + [-2.0294e-03, 8.3618e-03, -5.5237e-03, ..., 6.8054e-03, + 1.2451e-02, -7.4463e-03], + [ 4.2419e-03, -9.6436e-03, -1.7700e-03, ..., 5.3883e-05, + -5.3101e-03, 1.6022e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.mlp.layer2.weight': tensor([[-0.0023, 0.0052, 0.0056, ..., 0.0024, 0.0047, -0.0038], + [-0.0092, 0.0043, 0.0011, ..., -0.0048, 0.0008, 0.0156], + [ 0.0038, 0.0040, -0.0080, ..., 0.0023, -0.0068, 0.0004], + ..., + [ 0.0035, -0.0021, 0.0041, ..., 0.0055, -0.0041, 0.0023], + [-0.0184, 0.0021, 0.0183, ..., -0.0038, 0.0152, -0.0055], + [ 0.0076, -0.0065, -0.0046, ..., 0.0020, 0.0037, -0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0496, -0.0488, -0.0474, ..., 0.0031, 0.0032, 0.0032], + [ 0.0510, 0.0500, 0.0486, ..., -0.0025, -0.0026, -0.0027], + [ 0.0503, 0.0496, 0.0486, ..., -0.0027, -0.0027, -0.0028], + ..., + [ 0.0483, 0.0476, 0.0459, ..., -0.0024, -0.0025, -0.0025], + [ 0.0449, 0.0439, 0.0425, ..., -0.0022, -0.0023, -0.0022], + [ 0.0498, 0.0486, 0.0476, ..., -0.0031, -0.0032, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0046, 0.0041, 0.0043, ..., 0.0056, 0.0057, 0.0046], + [-0.0015, 0.0024, 0.0016, ..., 0.0019, 0.0019, 0.0014], + [ 0.0006, -0.0019, -0.0016, ..., 0.0003, 0.0020, -0.0006], + ..., + [-0.0371, 0.0469, 0.0457, ..., 0.0430, 0.0391, 0.0378], + [-0.0066, 0.0071, 0.0073, ..., 0.0052, 0.0022, 0.0069], + [ 0.0089, -0.0101, -0.0099, ..., -0.0090, -0.0086, -0.0090]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_cross_attn.1.weight': tensor([[-6.6223e-03, -6.2561e-03, -5.4626e-03, ..., -6.3324e-04, + -5.6076e-04, -7.3242e-04], + [-3.2715e-02, -3.1250e-02, -3.1128e-02, ..., 2.0790e-04, + 2.1458e-04, 2.2507e-04], + [ 3.8330e-02, 3.7109e-02, 3.6621e-02, ..., -1.0395e-04, + -8.5831e-05, -1.0967e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1250e-02, ..., -2.8801e-04, + -2.2316e-04, -2.5940e-04], + [ 4.7607e-02, 4.6143e-02, 4.5166e-02, ..., -1.9455e-04, + -1.3447e-04, -1.9360e-04], + [ 3.4424e-02, 3.2959e-02, 3.2715e-02, ..., -1.8358e-05, + 5.8889e-05, 9.4771e-06]], dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_cross_attn.2.weight': tensor([[-1.4526e-02, -4.3335e-03, 3.7079e-03, ..., 4.5776e-03, + 4.6082e-03, 3.4637e-03], + [ 7.0190e-04, -4.8828e-04, 2.1973e-03, ..., 5.8746e-04, + 3.2654e-03, 2.2430e-03], + [ 4.6082e-03, -3.8605e-03, 3.8605e-03, ..., 3.8452e-03, + -7.1049e-05, 3.8605e-03], + ..., + [ 6.1340e-03, -5.5847e-03, 2.4292e-02, ..., 5.5847e-03, + 2.7100e-02, 1.9165e-02], + [ 9.7046e-03, 5.7983e-03, 1.1963e-02, ..., -5.6152e-03, + 1.3046e-03, 1.2085e-02], + [-1.3245e-02, -4.3640e-03, 5.9891e-04, ..., 4.1809e-03, + -1.2817e-03, 1.7262e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_mlp.1.weight': tensor([[-0.0072, -0.0069, -0.0063, ..., 0.0005, 0.0007, 0.0007], + [ 0.0537, 0.0520, 0.0505, ..., -0.0024, -0.0027, -0.0027], + [ 0.0425, 0.0415, 0.0405, ..., -0.0035, -0.0038, -0.0037], + ..., + [ 0.0874, 0.0850, 0.0835, ..., -0.0009, -0.0011, -0.0011], + [ 0.0054, 0.0049, 0.0030, ..., -0.0005, -0.0006, -0.0005], + [ 0.0530, 0.0518, 0.0503, ..., -0.0012, -0.0012, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_mlp.2.weight': tensor([[-0.0023, 0.0044, 0.0044, ..., 0.0054, 0.0023, 0.0067], + [-0.0008, 0.0017, 0.0017, ..., 0.0037, 0.0008, -0.0004], + [ 0.0019, 0.0015, 0.0017, ..., 0.0017, -0.0018, -0.0018], + ..., + [-0.0001, 0.0586, 0.0698, ..., 0.0488, 0.0009, 0.0188], + [ 0.0099, 0.0119, 0.0101, ..., 0.0149, -0.0098, 0.0068], + [-0.0082, 0.0038, 0.0038, ..., 0.0038, 0.0081, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.before_proj.weight': tensor([[ 5.5730e-06, -2.1362e-03, 2.8534e-03, ..., -4.4556e-03, + -2.5330e-03, -1.2894e-03], + [ 3.4485e-03, 9.2163e-03, -1.2589e-03, ..., 1.5335e-03, + 3.1586e-03, -3.7994e-03], + [-3.1586e-03, -3.9816e-05, 1.6098e-03, ..., -4.7302e-04, + 4.9210e-04, -1.7624e-03], + ..., + [-5.6839e-04, -1.8387e-03, -3.8757e-03, ..., 9.8419e-04, + 1.6251e-03, -1.9836e-03], + [-8.8215e-05, 6.7749e-03, 1.5411e-03, ..., 2.6131e-04, + 6.3782e-03, 2.7771e-03], + [-6.0120e-03, -5.4932e-03, 1.7700e-03, ..., 6.0654e-04, + 1.2817e-03, -2.1057e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.before_proj.bias': tensor([-0.0082, 0.0026, 0.0023, ..., -0.0035, -0.0018, 0.0038], + dtype=torch.bfloat16), 'net.control_blocks_1.0.after_proj.weight': tensor([[-0.0084, 0.0053, 0.0074, ..., -0.0014, 0.0119, 0.0034], + [-0.0170, 0.0038, -0.0052, ..., 0.0042, -0.0137, -0.0132], + [-0.0028, 0.0058, -0.0018, ..., -0.0056, 0.0006, -0.0009], + ..., + [-0.0042, 0.0040, -0.0036, ..., -0.0186, 0.0045, 0.0073], + [ 0.0078, -0.0033, -0.0013, ..., 0.0085, -0.0168, -0.0064], + [ 0.0062, -0.0062, -0.0072, ..., -0.0015, -0.0031, 0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.after_proj.bias': tensor([ 0.0045, -0.0116, -0.0005, ..., -0.0031, 0.0062, 0.0292], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_proj.weight': tensor([[-3.4424e-02, 9.1553e-03, -9.8267e-03, ..., 1.7334e-02, + 9.5749e-04, -9.5825e-03], + [ 1.3733e-02, -1.9165e-02, 4.2725e-03, ..., -2.0874e-02, + 9.9487e-03, -1.8799e-02], + [-5.3406e-03, -3.6163e-03, -5.1575e-03, ..., -5.1880e-03, + -3.4943e-03, 1.9531e-03], + ..., + [ 2.7924e-03, 3.7384e-03, 2.3193e-03, ..., 4.3030e-03, + 4.5471e-03, 6.9141e-05], + [-3.3264e-03, -3.7231e-03, -1.6403e-03, ..., -3.5706e-03, + 4.1389e-04, -1.7242e-03], + [ 5.6458e-03, -3.7842e-03, 3.2654e-03, ..., -1.7090e-03, + -2.4872e-03, 3.5248e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_norm.weight': tensor([ 2.4512e-01, 8.0859e-01, 1.5430e-01, 1.0469e+00, 8.6719e-01, + 1.5122e-34, 1.9861e-34, -1.9410e-34, 7.5984e-35, 5.8487e-07, + 1.0909e-34, 5.4543e-35, 2.5203e-35, -1.4294e-34, -1.6476e-34, + 9.7049e-35, 1.7676e-01, 2.2888e-04, 2.3079e-04, 9.5215e-03, + 3.4142e-04, 7.1411e-03, 4.1406e-01, 1.0703e+00, 1.1406e+00, + 4.1797e-01, 5.8838e-02, 5.2734e-01, -1.7822e-02, 1.6777e-34, + -4.0324e-34, 1.9034e-34, -3.2806e-03, 9.2697e-04, 4.7684e-04, + 3.9978e-03, 5.1498e-04, -1.5422e-34, -4.9277e-35, 1.3692e-34, + 1.3316e-34, -1.8161e-08, 1.3466e-34, 1.1094e+00, 5.5078e-01, + 4.0039e-01, 3.0664e-01, 1.2422e+00, 1.0547e-01, 1.2360e-03, + 1.1780e-02, 1.9043e-01, 2.6855e-03, -6.9965e-35, -1.6175e-34, + -1.8996e-35, 1.0193e-02, 2.7195e-07, 1.2488e-34, -8.4636e-35, + 3.8556e-35, 3.8744e-35, 1.0608e-34, -2.7836e-35, 5.5078e-01, + 2.7539e-01, 6.7578e-01, 3.6328e-01, 1.3574e-01, -9.5544e-35, + -6.8009e-34, 1.5046e-34, 4.5139e-35, 3.7998e-07, 1.1661e-34, + 1.9259e-34, 2.7836e-35, 7.5608e-35, -1.4219e-34, -1.3090e-34, + -9.3994e-03, 7.6675e-04, 4.4823e-04, 1.6632e-03, 9.3079e-04, + 4.8096e-02, 1.1016e+00, 4.3750e-01, 4.9023e-01, 1.1406e+00, + 2.8516e-01, 2.0117e-01, 6.5308e-03, -5.4769e-34, 2.6933e-34, + 2.0162e-34, 2.4986e-04, -2.4033e-04, 4.7874e-04, 4.5166e-03, + 8.5068e-04, 2.6519e-35, -4.1754e-35, 1.9259e-34, 1.1962e-34, + 1.2293e-07, -1.9259e-34, 3.0859e-01, 1.0625e+00, 1.1250e+00, + 1.2031e+00, 1.7773e-01, 7.6953e-01, 9.1553e-03, -1.9684e-03, + 2.1973e-03, 3.9978e-03, 2.8438e-34, -1.3015e-34, 3.3704e-34, + 9.9945e-04, 1.3828e-05, 6.9589e-35, 3.0394e-34, -1.1285e-35, + -2.1065e-34, 2.1535e-35, 6.3947e-35], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.self_attn.k_proj.weight': tensor([[ 0.0067, -0.0266, 0.0039, ..., -0.0156, -0.0106, -0.0172], + [ 0.0134, -0.0278, 0.0018, ..., -0.0325, 0.0129, -0.0195], + [ 0.0085, 0.0120, -0.0021, ..., -0.0072, -0.0024, -0.0022], + ..., + [ 0.0030, 0.0064, -0.0003, ..., 0.0005, 0.0013, -0.0032], + [-0.0025, -0.0030, -0.0014, ..., -0.0003, 0.0021, -0.0052], + [-0.0003, -0.0024, 0.0031, ..., -0.0048, 0.0014, -0.0027]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.k_norm.weight': tensor([ 4.0234e-01, 8.4766e-01, 2.1777e-01, 1.0625e+00, 8.6719e-01, + 5.9057e-35, 6.3947e-35, 6.7332e-35, -7.4855e-35, -2.3842e-07, + -1.0833e-34, -4.2882e-35, 4.4387e-35, -1.0006e-34, -1.4520e-34, + -1.3843e-34, 1.7676e-01, 3.7766e-04, 6.3896e-05, 4.6387e-02, + 3.3760e-04, -7.2937e-03, 5.1562e-01, 1.0703e+00, 1.1484e+00, + 3.4766e-01, 1.8066e-01, 6.1719e-01, -1.9653e-02, 2.7384e-34, + -4.8524e-35, 1.4896e-34, 3.4027e-03, 9.7656e-04, -1.9169e-04, + 4.6143e-02, 5.2643e-04, -5.0781e-35, -2.4262e-35, 8.5388e-35, + -7.7865e-35, -2.6659e-08, 1.6927e-34, 1.1094e+00, 4.1602e-01, + 1.2012e-01, -2.0605e-01, 1.2266e+00, 1.1377e-01, 1.6479e-03, + 9.2163e-03, 1.2354e-01, 2.1057e-03, 1.0457e-34, 2.0463e-34, + 4.3634e-35, 1.0132e-02, -3.3760e-08, -1.5723e-34, -7.8241e-35, + -1.2413e-35, -6.1314e-35, -1.4294e-34, -2.2569e-35, 4.9023e-01, + 3.3984e-01, 6.7969e-01, 2.5000e-01, 6.2500e-02, 1.5122e-34, + -1.0713e-33, 2.3623e-34, 2.3510e-35, -7.1526e-07, 2.4225e-34, + 2.3623e-34, 1.9184e-35, 8.0874e-35, 2.1535e-35, 1.0382e-34, + 1.7578e-02, 8.2779e-04, 1.6332e-05, 1.4267e-03, 9.3079e-04, + 4.7852e-02, 1.0938e+00, 3.0273e-01, 3.2617e-01, 1.1484e+00, + 2.3926e-01, 1.9434e-01, -2.1362e-04, -1.4520e-34, 2.5127e-34, + -2.9792e-34, -7.9155e-05, 4.1580e-04, 1.8978e-04, 4.5410e-02, + 8.5068e-04, 3.7428e-35, -4.5515e-35, -1.9259e-34, 8.5388e-35, + -1.2480e-07, -2.1516e-34, 3.5156e-01, 1.0781e+00, 1.1328e+00, + 1.2031e+00, -1.1182e-01, 7.8516e-01, 9.2163e-03, 6.6833e-03, + -1.6016e-01, -3.6011e-03, 4.3634e-34, -1.0608e-34, 4.9051e-34, + -1.6403e-03, 1.3828e-05, 2.9716e-35, 1.9184e-34, -3.7616e-35, + -1.0382e-34, -8.4259e-35, -4.7208e-35], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.self_attn.v_proj.weight': tensor([[ 0.0094, 0.0021, -0.0027, ..., -0.0101, -0.0006, 0.0021], + [ 0.0112, -0.0018, 0.0096, ..., 0.0132, -0.0034, -0.0018], + [ 0.0030, 0.0090, -0.0054, ..., 0.0119, 0.0067, -0.0090], + ..., + [ 0.0101, 0.0050, 0.0096, ..., 0.0025, -0.0022, -0.0150], + [ 0.0035, 0.0030, 0.0087, ..., 0.0302, -0.0042, 0.0026], + [-0.0038, -0.0031, -0.0037, ..., -0.0071, 0.0027, 0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.output_proj.weight': tensor([[-0.0118, -0.0044, -0.0154, ..., 0.0201, -0.0093, -0.0011], + [ 0.0154, 0.0047, 0.0190, ..., 0.0036, -0.0079, 0.0216], + [-0.0004, 0.0034, 0.0029, ..., -0.0008, 0.0037, 0.0030], + ..., + [-0.0063, 0.0133, 0.0054, ..., -0.0106, -0.0228, 0.0036], + [ 0.0079, -0.0027, -0.0053, ..., -0.0088, -0.0022, 0.0038], + [ 0.0028, 0.0053, -0.0009, ..., 0.0108, 0.0050, 0.0019]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_proj.weight': tensor([[-0.0126, 0.0129, 0.0119, ..., 0.0013, 0.0060, -0.0038], + [ 0.0101, -0.0017, 0.0041, ..., 0.0045, -0.0068, -0.0172], + [ 0.0040, 0.0110, -0.0029, ..., 0.0031, 0.0066, -0.0067], + ..., + [ 0.0084, -0.0129, 0.0032, ..., -0.0044, -0.0045, 0.0088], + [-0.0099, 0.0049, 0.0044, ..., 0.0073, 0.0020, 0.0093], + [ 0.0398, -0.0062, -0.0023, ..., 0.0098, 0.0096, -0.0177]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_norm.weight': tensor([ 2.2217e-02, 4.4922e-01, 2.2583e-02, 9.9487e-03, 4.8218e-03, + 8.4639e-06, 7.0572e-04, 5.1562e-01, 7.6675e-04, 3.9062e-01, + 2.9785e-02, 1.9653e-02, 8.4229e-03, 1.5015e-02, 1.8921e-02, + 1.4648e-02, 8.5449e-03, -9.6798e-05, 6.1512e-05, 3.1738e-02, + 2.6172e-01, 2.5635e-02, 5.4016e-03, 1.2512e-02, 6.2561e-03, + 1.2302e-04, -2.0266e-05, 4.8584e-02, 7.5102e-06, 3.7109e-01, + 2.0790e-04, 7.8082e-06, 9.6436e-03, -3.2336e-06, 1.0559e-02, + 7.2021e-03, 6.2466e-05, 3.9844e-01, 1.4343e-02, 6.5613e-03, + 7.0801e-02, 6.4392e-03, 1.0986e-02, 4.5471e-03, 1.2817e-02, + 8.6060e-03, 6.3171e-03, 1.7822e-02, 2.4902e-02, 1.3062e-02, + 1.4160e-02, 4.0234e-01, 1.1169e-02, 1.9897e-02, 3.1982e-02, + 1.3351e-03, 9.6680e-02, 1.3611e-02, 2.0264e-02, 1.4954e-02, + -5.2750e-06, 1.9043e-02, 1.5625e-02, 1.3053e-05, 1.2268e-02, + 8.2397e-03, 1.4496e-04, 8.3923e-04, 1.6357e-02, 7.8125e-02, + 1.2131e-03, 1.3123e-02, 4.8828e-02, 3.9795e-02, 1.6327e-03, + 1.8433e-02, -1.6212e-05, 2.5269e-02, 2.5635e-02, 2.5269e-02, + 6.8970e-03, 1.8555e-02, 3.4424e-02, -3.5286e-05, 2.1240e-02, + 3.3569e-03, 1.5030e-03, 4.4189e-02, 6.3782e-03, 1.4267e-03, + 1.6602e-02, 2.2030e-04, 7.4768e-03, 2.0020e-02, 9.0942e-03, + 5.9128e-05, 1.7456e-02, 4.4189e-02, 1.5991e-02, 4.2236e-02, + 3.8330e-02, 2.7588e-02, 1.2878e-02, 9.3460e-05, -3.5286e-05, + 5.5847e-03, 1.1414e-02, 2.1484e-02, 1.7456e-02, 2.5269e-02, + 7.3547e-03, 6.4087e-03, 3.5400e-02, 4.9023e-01, 3.0212e-03, + 1.6724e-02, 6.8054e-03, 1.1292e-02, 7.5989e-03, 1.0547e-01, + 1.4648e-03, 1.3733e-03, -9.6560e-06, 1.2634e-02, 5.9204e-03, + 2.1851e-02, 2.7466e-02, 1.6846e-02], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_proj.weight': tensor([[-0.0032, -0.0033, -0.0042, ..., -0.0037, -0.0011, 0.0017], + [ 0.0116, -0.0035, 0.0184, ..., 0.0208, 0.0198, 0.0116], + [-0.0008, 0.0036, 0.0039, ..., 0.0021, 0.0052, -0.0018], + ..., + [-0.0070, -0.0049, -0.0031, ..., -0.0012, 0.0021, -0.0016], + [-0.0072, -0.0053, -0.0142, ..., 0.0007, 0.0086, -0.0025], + [ 0.0007, -0.0004, -0.0055, ..., -0.0012, 0.0014, -0.0043]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_norm.weight': tensor([ 2.2827e-02, 4.4922e-01, 2.2461e-02, 9.9487e-03, 4.8523e-03, + -6.1095e-06, 7.2098e-04, 5.1562e-01, 8.0109e-04, 3.9062e-01, + 3.0518e-02, 1.9653e-02, 8.4229e-03, 1.5015e-02, 1.8921e-02, + 1.4526e-02, 9.0942e-03, -9.7275e-05, 5.2214e-05, 3.1982e-02, + 2.6172e-01, 2.4902e-02, 5.4321e-03, 1.2756e-02, 6.1951e-03, + 1.2493e-04, -1.8597e-05, 4.8096e-02, 6.1989e-06, 3.7109e-01, + 2.0885e-04, 9.1791e-06, 9.3384e-03, -1.7643e-05, 1.0559e-02, + 7.2632e-03, 6.2466e-05, 3.9844e-01, 1.4404e-02, 6.5002e-03, + 6.9824e-02, 6.4087e-03, 1.0803e-02, 4.5471e-03, 1.2756e-02, + 8.5449e-03, 6.4087e-03, 1.7700e-02, 2.4780e-02, 1.3306e-02, + 1.4160e-02, 4.0234e-01, 1.1169e-02, 1.9531e-02, 3.1982e-02, + 1.3275e-03, 9.6680e-02, 1.3611e-02, 1.9897e-02, 1.5076e-02, + -7.5102e-06, 1.9043e-02, 1.5625e-02, 1.3053e-05, 1.2207e-02, + 8.1177e-03, 1.4400e-04, 8.3542e-04, 1.6357e-02, 7.8125e-02, + 1.1597e-03, 1.3123e-02, 4.9072e-02, 3.9795e-02, 1.6632e-03, + 1.8555e-02, -2.3246e-05, 2.5024e-02, 2.5757e-02, 2.5269e-02, + 6.8970e-03, 1.8433e-02, 3.3936e-02, -3.6001e-05, 2.1118e-02, + 3.2501e-03, 1.4801e-03, 4.3945e-02, 6.6528e-03, 1.7242e-03, + 1.6968e-02, 2.5558e-04, 7.4158e-03, 1.9897e-02, 8.9111e-03, + 5.9128e-05, 1.6968e-02, 4.3701e-02, 1.5991e-02, 4.1992e-02, + 3.9062e-02, 2.7588e-02, 1.3000e-02, 9.0122e-05, -3.5524e-05, + 5.5542e-03, 1.1475e-02, 2.1240e-02, 1.7456e-02, 2.5146e-02, + 7.3547e-03, 6.1646e-03, 3.4912e-02, 4.8828e-01, 3.0670e-03, + 1.6724e-02, 6.8054e-03, 1.1169e-02, 7.7820e-03, 1.0547e-01, + 1.5488e-03, 1.3199e-03, -5.1856e-06, 1.3000e-02, 5.9814e-03, + 2.1973e-02, 2.7466e-02, 1.6724e-02], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.v_proj.weight': tensor([[-0.0178, -0.0001, -0.0044, ..., 0.0012, -0.0193, -0.0247], + [-0.0055, 0.0055, 0.0051, ..., 0.0037, 0.0095, 0.0070], + [-0.0400, -0.0034, -0.0039, ..., -0.0058, -0.0315, -0.0082], + ..., + [-0.0417, 0.0030, -0.0164, ..., 0.0028, 0.0205, 0.0055], + [ 0.0082, 0.0029, -0.0052, ..., 0.0016, 0.0115, 0.0044], + [ 0.0112, 0.0037, -0.0070, ..., 0.0033, 0.0131, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.output_proj.weight': tensor([[ 0.0064, 0.0042, 0.0162, ..., -0.0038, 0.0042, 0.0040], + [-0.0072, 0.0031, -0.0103, ..., -0.0123, 0.0031, 0.0117], + [-0.0011, 0.0020, -0.0045, ..., 0.0019, 0.0011, -0.0018], + ..., + [-0.0021, -0.0033, 0.0043, ..., -0.0018, 0.0036, 0.0046], + [ 0.0087, 0.0232, 0.0022, ..., -0.0349, 0.0075, 0.0289], + [ 0.0039, 0.0022, -0.0156, ..., -0.0105, 0.0062, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_img.weight': tensor([[ 0.0079, -0.0041, 0.0430, ..., -0.0327, 0.0048, 0.0144], + [-0.0278, 0.0280, 0.0249, ..., -0.0077, -0.0378, 0.0271], + [ 0.0139, -0.0119, 0.0312, ..., -0.0072, 0.0253, -0.0111], + ..., + [ 0.0245, -0.0152, 0.0043, ..., -0.0452, 0.0216, -0.0209], + [-0.0089, 0.0076, -0.0471, ..., -0.0107, 0.0157, -0.0083], + [ 0.0074, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0027, -0.0122, -0.0108], + [ 0.0135, 0.0048, -0.0188, ..., 0.0156, 0.0074, -0.0045], + [-0.0306, -0.0021, -0.0272, ..., 0.0103, 0.0337, 0.0017], + ..., + [-0.0034, -0.0042, 0.0049, ..., 0.0208, -0.0107, 0.0065], + [ 0.0193, 0.0210, 0.0122, ..., -0.0145, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0106, -0.0132, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img.weight': tensor([[ 0.0038, -0.0043, 0.0173, ..., 0.0083, 0.0356, 0.0217], + [-0.0498, -0.0033, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0108, 0.0043, -0.0139, ..., 0.0007, 0.0269, 0.0072], + ..., + [-0.0043, 0.0126, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0302, 0.0080, -0.0162, ..., -0.0050, -0.0074, -0.0118], + [ 0.0254, 0.0130, -0.0425, ..., -0.0264, 0.0138, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.mlp.layer1.weight': tensor([[ 0.0022, -0.0042, 0.0031, ..., 0.0037, 0.0016, -0.0054], + [-0.0031, 0.0053, 0.0071, ..., -0.0023, -0.0075, 0.0022], + [ 0.0026, 0.0027, 0.0036, ..., -0.0020, -0.0065, 0.0103], + ..., + [ 0.0281, 0.0128, -0.0095, ..., 0.0020, -0.0066, 0.0111], + [-0.0449, 0.0030, -0.0078, ..., 0.0183, 0.0075, -0.0037], + [ 0.0094, -0.0079, 0.0070, ..., -0.0075, 0.0014, -0.0247]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.mlp.layer2.weight': tensor([[-0.0096, -0.0112, 0.0197, ..., 0.0036, -0.0042, -0.0118], + [-0.0065, 0.0038, -0.0004, ..., -0.0036, -0.0001, -0.0032], + [ 0.0049, 0.0027, -0.0094, ..., -0.0126, 0.0036, 0.0025], + ..., + [ 0.0063, -0.0124, 0.0052, ..., -0.0036, -0.0082, 0.0097], + [ 0.0051, 0.0172, 0.0078, ..., 0.0063, 0.0067, 0.0023], + [ 0.0024, -0.0042, 0.0013, ..., -0.0013, -0.0005, 0.0079]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0410, -0.0398, -0.0393, ..., 0.0050, 0.0053, 0.0052], + [ 0.0327, 0.0315, 0.0305, ..., -0.0039, -0.0038, -0.0039], + [-0.0449, -0.0435, -0.0430, ..., 0.0051, 0.0052, 0.0052], + ..., + [-0.0454, -0.0442, -0.0430, ..., 0.0034, 0.0035, 0.0036], + [-0.0403, -0.0396, -0.0388, ..., 0.0037, 0.0039, 0.0039], + [-0.0231, -0.0222, -0.0211, ..., 0.0014, 0.0014, 0.0015]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_self_attn.2.weight': tensor([[ 6.8359e-03, -1.9302e-03, 5.8594e-03, ..., 6.8054e-03, + 6.8970e-03, 2.0142e-03], + [-1.6174e-03, 1.6403e-03, -1.7319e-03, ..., -2.3956e-03, + -1.7853e-03, -7.0572e-04], + [-1.0864e-02, -5.0735e-04, -9.8877e-03, ..., -1.1292e-02, + -1.0803e-02, 2.0862e-05], + ..., + [-1.8433e-02, -4.3678e-04, -2.4414e-02, ..., -2.6611e-02, + -2.5757e-02, 8.3542e-04], + [-1.0986e-02, -6.4697e-03, -1.1780e-02, ..., -1.1780e-02, + -1.0376e-02, 7.3242e-03], + [ 1.7395e-03, -6.0730e-03, 2.8534e-03, ..., 2.6550e-03, + 2.0447e-03, 5.0049e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5176e-02, 5.3711e-02, 5.3223e-02, ..., -1.0872e-04, + -1.5354e-04, -1.0729e-04], + [ 4.2114e-03, 4.1504e-03, 4.1504e-03, ..., -7.5340e-05, + -1.5163e-04, -2.6703e-04], + [-1.7700e-02, -1.6968e-02, -1.6235e-02, ..., -6.6757e-04, + -6.6757e-04, -6.7902e-04], + ..., + [-2.6001e-02, -2.6001e-02, -2.5024e-02, ..., -4.2343e-04, + -3.8147e-04, -3.4332e-04], + [-2.7100e-02, -2.6733e-02, -2.5635e-02, ..., -2.1839e-04, + -1.7262e-04, -2.2602e-04], + [ 5.4199e-02, 5.3223e-02, 5.2246e-02, ..., -1.7262e-04, + -5.4836e-05, -2.0218e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_cross_attn.2.weight': tensor([[-8.9111e-03, -4.9133e-03, -1.1353e-02, ..., -6.1646e-03, + 9.2163e-03, -9.5825e-03], + [ 9.6436e-03, -1.9150e-03, -3.1128e-03, ..., 2.9755e-03, + -8.4229e-03, 1.0071e-02], + [ 5.7983e-03, 1.8215e-04, -2.6941e-05, ..., 3.1281e-03, + -4.3335e-03, 5.3101e-03], + ..., + [ 6.0547e-02, 7.0190e-03, -4.3030e-03, ..., 5.7678e-03, + -1.6113e-02, 6.0059e-02], + [ 1.0620e-02, -4.6082e-03, -8.7891e-03, ..., 4.1504e-03, + -1.5625e-02, 1.1230e-02], + [-3.8605e-03, 9.1553e-03, -1.5503e-02, ..., 1.0376e-02, + 7.3242e-03, -3.8452e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_mlp.1.weight': tensor([[-2.8687e-02, -2.8564e-02, -2.8198e-02, ..., 3.9816e-05, + -4.6194e-06, 1.8477e-05], + [ 3.6621e-02, 3.6133e-02, 3.5889e-02, ..., -3.5248e-03, + -3.5553e-03, -3.4790e-03], + [ 3.9795e-02, 3.9062e-02, 3.8574e-02, ..., -3.6621e-03, + -3.6621e-03, -3.7079e-03], + ..., + [ 4.1016e-02, 3.9795e-02, 3.9551e-02, ..., -1.8616e-03, + -1.8845e-03, -1.8234e-03], + [ 1.9653e-02, 2.0264e-02, 2.0264e-02, ..., 1.6937e-03, + 1.6708e-03, 1.5640e-03], + [ 3.4912e-02, 3.4912e-02, 3.4668e-02, ..., -2.2583e-03, + -2.3193e-03, -2.3346e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_mlp.2.weight': tensor([[-0.0033, 0.0037, -0.0026, ..., 0.0036, 0.0035, 0.0038], + [ 0.0029, -0.0019, -0.0007, ..., -0.0013, -0.0033, -0.0018], + [ 0.0055, -0.0088, -0.0107, ..., -0.0078, -0.0045, -0.0078], + ..., + [ 0.0024, -0.0120, -0.0199, ..., -0.0101, 0.0016, -0.0090], + [ 0.0038, -0.0038, -0.0087, ..., -0.0038, -0.0038, -0.0038], + [ 0.0086, -0.0053, -0.0040, ..., -0.0063, -0.0093, -0.0060]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.after_proj.weight': tensor([[-3.2227e-02, 6.6528e-03, 3.7384e-03, ..., -1.4465e-02, + 7.3242e-03, -5.8899e-03], + [ 9.6436e-03, -6.9141e-05, 6.8054e-03, ..., 1.1902e-02, + 6.5918e-03, 1.8539e-03], + [ 1.4771e-02, -7.2021e-03, -1.2085e-02, ..., -4.2419e-03, + 5.5847e-03, 4.0894e-03], + ..., + [-7.9956e-03, 6.7139e-03, 4.1504e-03, ..., 1.5198e-02, + 6.6223e-03, 6.1340e-03], + [-1.2024e-02, -2.3499e-03, 5.7983e-03, ..., 5.2795e-03, + 6.1646e-03, 6.5613e-03], + [ 5.5847e-03, -7.5684e-03, 5.6458e-03, ..., 5.3406e-03, + -6.8359e-03, -1.2894e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.after_proj.bias': tensor([-0.0041, -0.0006, 0.0020, ..., -0.0030, -0.0014, -0.0011], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_proj.weight': tensor([[ 0.0452, -0.0189, -0.0077, ..., 0.0088, -0.0152, 0.0352], + [-0.0103, -0.0201, -0.0019, ..., -0.0052, 0.0085, -0.0058], + [ 0.0259, 0.0054, -0.0006, ..., 0.0060, -0.0057, 0.0159], + ..., + [-0.0143, -0.0048, 0.0012, ..., 0.0034, -0.0030, -0.0065], + [ 0.0223, 0.0277, 0.0019, ..., 0.0176, -0.0047, 0.0086], + [-0.0100, 0.0103, 0.0188, ..., -0.0136, -0.0041, -0.0189]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_norm.weight': tensor([ 1.4844e+00, 3.2617e-01, 4.1602e-01, 3.4570e-01, 1.3750e+00, + 2.7466e-04, 1.0437e-02, 5.0962e-06, 1.9775e-02, 5.3024e-04, + -9.7789e-08, -2.5705e-07, 1.1484e+00, 3.9978e-03, -3.8369e-13, + -1.6479e-03, -1.9379e-03, -2.3499e-03, 3.0396e-02, 1.5869e-02, + 1.5820e-01, -8.8692e-05, 4.2578e-01, 1.3594e+00, 5.0000e-01, + 5.5078e-01, 3.8477e-01, 1.6699e-01, 2.5513e-02, -2.5482e-03, + 1.3611e-02, -7.2098e-04, 1.4160e-02, 5.8899e-03, -1.2512e-03, + 3.4180e-02, 1.2988e-01, 5.9891e-04, -2.2430e-03, -4.0829e-06, + -1.2589e-04, 1.5182e-03, 2.2888e-03, 1.1562e+00, 9.0332e-02, + 1.3828e+00, 5.8203e-01, 2.0117e-01, -3.6865e-02, 4.9744e-03, + -2.3651e-03, 3.0273e-02, 1.3086e-01, 5.3048e-06, -7.1526e-05, + -1.1176e-06, -2.0981e-04, -6.3419e-05, -4.4678e-02, 4.8021e-09, + 3.5156e-02, -1.7262e-04, -2.9541e-02, -2.0020e-02, 6.0156e-01, + 1.4297e+00, 1.3281e+00, 1.3984e+00, 2.9688e-01, -1.6846e-02, + 1.3245e-02, 7.3016e-06, 2.1118e-02, -1.3367e-02, 1.4808e-07, + -2.3395e-06, 5.8105e-02, 2.7618e-03, 2.0073e-13, 8.1250e-01, + 1.0625e+00, 1.3000e-02, 1.9550e-04, 3.1982e-02, 1.1230e-01, + 9.8267e-03, 1.3359e+00, 3.1250e-01, 1.3672e+00, 1.4844e+00, + 2.2266e-01, 4.8047e-01, 4.7852e-02, -2.0630e-02, -5.2185e-03, + -6.7234e-05, -9.0332e-03, 1.4832e-02, -2.1515e-03, 3.3447e-02, + 1.2158e-01, 6.1989e-05, 1.1562e+00, 6.1095e-07, -4.6015e-05, + -3.7842e-03, 3.0100e-06, 7.1484e-01, 1.4219e+00, 1.3672e+00, + 1.5781e+00, 5.2734e-01, 2.7275e-04, 5.5542e-03, 6.3171e-03, + 3.1250e-02, 1.3770e-01, 1.4365e-05, -2.9504e-06, -6.2957e-07, + -6.6757e-05, -1.0452e-03, 1.1250e+00, 4.5262e-07, 2.8442e-02, + -1.5378e-05, 7.8125e-01, 1.1016e+00], dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.self_attn.k_proj.weight': tensor([[-0.0225, 0.0005, 0.0068, ..., 0.0205, 0.0119, 0.0415], + [-0.0079, 0.0108, -0.0090, ..., -0.0039, -0.0042, -0.0036], + [-0.0143, 0.0108, -0.0109, ..., 0.0058, 0.0182, 0.0007], + ..., + [-0.0030, -0.0069, 0.0044, ..., 0.0017, 0.0044, 0.0078], + [-0.0130, -0.0059, 0.0067, ..., 0.0051, 0.0056, 0.0075], + [ 0.0010, -0.0042, 0.0038, ..., -0.0051, -0.0101, 0.0044]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.k_norm.weight': tensor([ 1.5156e+00, 7.0703e-01, 3.8867e-01, 2.8711e-01, 1.3828e+00, + -7.5989e-03, 2.1484e-02, -4.3511e-06, 2.9663e-02, -1.7471e-03, + -1.0943e-07, -2.4959e-07, 1.1484e+00, 3.7537e-03, -3.7126e-13, + -1.6602e-02, -9.8419e-04, 8.0078e-02, 3.2227e-02, -3.7384e-03, + 5.3101e-03, 9.2030e-05, 5.8594e-01, 1.2344e+00, 8.6426e-02, + 5.4297e-01, 4.4141e-01, 1.6406e-01, 4.8828e-03, -5.6763e-03, + 1.4709e-02, -7.8201e-04, 1.3306e-02, 4.9561e-02, 1.9226e-03, + -1.7014e-03, 8.1787e-03, 5.9128e-04, 4.0894e-03, 4.0829e-06, + -1.1444e-04, 1.4114e-03, 2.2583e-03, 1.2891e+00, 6.6797e-01, + 1.3125e+00, 5.8203e-01, 2.9492e-01, -4.8828e-03, 6.2012e-02, + 3.7231e-03, -6.1951e-03, 3.3722e-03, -1.6928e-05, 6.5804e-05, + 1.1697e-06, -2.0885e-04, -3.1710e-05, -9.8877e-03, -3.7835e-09, + 5.4932e-03, -1.6975e-04, -3.5645e-02, -1.4099e-02, 7.0312e-01, + 1.3359e+00, 1.3359e+00, 1.3828e+00, 2.5977e-01, 1.5259e-02, + -2.2125e-03, 8.3447e-06, -2.8687e-03, -1.3184e-02, -1.3318e-07, + 2.3246e-06, 5.6763e-03, 3.0365e-03, 3.0376e-13, 8.1250e-01, + 1.0625e+00, 6.7871e-02, 1.3733e-04, -4.4861e-03, -4.0894e-03, + 9.9487e-03, 1.2500e+00, 7.3047e-01, 1.4766e+00, 1.5312e+00, + 4.5410e-02, 4.8828e-01, -1.4160e-02, 2.1118e-02, 1.6251e-03, + -6.7234e-05, 1.3794e-02, 5.3223e-02, 1.0071e-02, -5.9204e-03, + 8.5449e-03, -3.3617e-05, 1.1562e+00, 6.2212e-07, -8.0585e-05, + -3.8300e-03, 4.7386e-06, 3.7891e-01, 1.2891e+00, 1.3672e+00, + 1.6953e+00, 5.3125e-01, -2.5024e-02, 8.5938e-02, 6.5613e-03, + 1.7334e-02, -5.0964e-03, 1.1146e-05, 1.3709e-05, 5.3644e-07, + -2.5749e-05, -1.0529e-03, 1.1250e+00, -5.0664e-07, -7.2021e-03, + 2.2411e-05, 7.8125e-01, 1.1016e+00], dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.self_attn.v_proj.weight': tensor([[-0.0078, 0.0123, -0.0081, ..., -0.0261, 0.0009, 0.0117], + [ 0.0104, -0.0057, -0.0150, ..., 0.0039, -0.0039, 0.0118], + [-0.0068, -0.0043, 0.0041, ..., 0.0190, -0.0097, -0.0103], + ..., + [ 0.0154, -0.0128, -0.0153, ..., -0.0254, -0.0073, 0.0041], + [ 0.0048, -0.0039, 0.0100, ..., 0.0019, 0.0038, -0.0142], + [-0.0115, 0.0057, -0.0239, ..., 0.0041, 0.0121, 0.0092]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.output_proj.weight': tensor([[ 0.0040, 0.0135, -0.0322, ..., 0.0049, -0.0205, 0.0128], + [ 0.0024, -0.0211, -0.0120, ..., 0.0063, 0.0154, -0.0089], + [-0.0041, 0.0023, 0.0042, ..., 0.0028, 0.0031, 0.0042], + ..., + [ 0.0118, 0.0081, -0.0229, ..., 0.0190, 0.0114, 0.0009], + [ 0.0064, 0.0043, 0.0041, ..., 0.0077, 0.0093, 0.0035], + [ 0.0062, 0.0047, -0.0041, ..., -0.0068, 0.0107, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_proj.weight': tensor([[-0.0069, 0.0082, 0.0011, ..., -0.0103, -0.0052, 0.0002], + [ 0.0091, 0.0072, 0.0038, ..., -0.0041, -0.0059, 0.0029], + [ 0.0051, 0.0062, -0.0024, ..., -0.0035, 0.0157, -0.0038], + ..., + [ 0.0027, 0.0070, 0.0029, ..., 0.0079, 0.0060, 0.0046], + [-0.0107, 0.0010, 0.0054, ..., 0.0030, -0.0099, 0.0085], + [ 0.0070, 0.0062, -0.0022, ..., -0.0051, 0.0134, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_norm.weight': tensor([3.5400e-03, 1.4551e-01, 1.3245e-02, 5.8594e-01, 2.1240e-02, 2.3315e-02, + 8.7891e-03, 4.8523e-03, 1.7456e-02, 1.2598e-01, 8.4839e-03, 5.5664e-02, + 2.7344e-01, 1.9165e-02, 1.8945e-01, 1.5320e-02, 6.7871e-02, 1.5442e-02, + 4.1504e-03, 1.8555e-02, 1.0559e-02, 1.8652e-01, 2.8711e-01, 1.9922e-01, + 2.3315e-02, 1.5918e-01, 6.6406e-02, 1.6357e-02, 4.0625e-01, 8.6670e-03, + 1.9775e-02, 5.1172e-01, 6.6895e-02, 3.4668e-02, 1.5991e-02, 1.5991e-02, + 7.5195e-02, 4.3359e-01, 5.3125e-01, 1.2451e-02, 8.6060e-03, 1.2109e-01, + 5.6152e-03, 5.2344e-01, 3.0273e-02, 5.0293e-02, 1.6235e-02, 3.8086e-01, + 7.6294e-03, 1.8555e-02, 8.2520e-02, 1.2451e-02, 1.8799e-02, 4.6484e-01, + 1.8768e-03, 2.6978e-02, 5.4932e-02, 8.0078e-02, 6.2256e-03, 1.5991e-02, + 4.8523e-03, 2.3926e-02, 1.0559e-02, 3.6133e-01, 1.6602e-02, 3.0670e-03, + 2.1118e-02, 5.9570e-02, 5.9814e-02, 1.3611e-02, 6.6528e-03, 6.0156e-01, + 3.7575e-04, 5.7031e-01, 1.3855e-02, 3.9673e-03, 1.6602e-02, 3.0664e-01, + 2.6733e-02, 3.3722e-03, 1.0986e-01, 5.2490e-03, 6.3477e-02, 9.1553e-03, + 1.9897e-02, 8.1787e-03, 1.2390e-02, 6.5430e-02, 2.0508e-01, 1.0315e-02, + 1.5564e-03, 1.8433e-02, 4.6387e-02, 2.4170e-02, 2.6245e-02, 1.5747e-02, + 1.7212e-02, 2.3193e-02, 6.3965e-02, 2.7344e-02, 4.3213e-02, 1.5198e-02, + 3.1738e-03, 6.5430e-02, 5.8594e-01, 7.6172e-02, 6.8359e-03, 4.8047e-01, + 1.3428e-02, 3.0762e-02, 1.7456e-02, 1.3000e-02, 1.4591e-04, 9.0332e-02, + 4.7852e-01, 2.4170e-02, 2.4292e-02, 5.7422e-01, 1.2158e-01, 2.6001e-02, + 1.9775e-02, 6.8750e-01, 1.5234e-01, 1.5869e-02, 1.4404e-02, 5.4932e-03, + 2.8442e-02, 1.1292e-02], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_proj.weight': tensor([[ 2.1057e-03, -3.3875e-03, 4.6692e-03, ..., 6.9427e-04, + -4.0283e-03, 4.7302e-03], + [-7.0190e-03, 2.6321e-04, -5.3711e-03, ..., 5.0964e-03, + -7.8125e-03, -5.4626e-03], + [-5.3711e-03, -1.6174e-03, -5.2490e-03, ..., 1.2451e-02, + -1.4526e-02, -3.6926e-03], + ..., + [ 6.9580e-03, 4.6082e-03, 2.5482e-03, ..., 2.9297e-03, + -1.8768e-03, -6.6223e-03], + [ 6.2866e-03, 2.9449e-03, 1.3504e-03, ..., -1.3275e-03, + 2.6398e-03, 4.4250e-03], + [ 6.9141e-05, 6.7902e-04, -1.2589e-03, ..., -6.3782e-03, + -6.1340e-03, 3.0823e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_norm.weight': tensor([ 3.5400e-03, 1.4551e-01, 1.3062e-02, 5.8594e-01, 2.0996e-02, + 2.3193e-02, 8.9722e-03, 4.7302e-03, 1.7700e-02, 1.2598e-01, + 8.4839e-03, 5.3955e-02, 2.7344e-01, 1.9287e-02, 1.8945e-01, + 1.6113e-02, 6.7383e-02, 1.5442e-02, 4.0894e-03, 1.8555e-02, + 1.0803e-02, 1.8750e-01, 2.8906e-01, 1.9922e-01, 2.3071e-02, + 1.5918e-01, 6.5430e-02, 1.5137e-02, 4.0820e-01, 8.6670e-03, + 2.0020e-02, 5.1172e-01, 6.6895e-02, 3.4424e-02, 1.6357e-02, + 1.5625e-02, 7.5195e-02, 4.3359e-01, 5.3125e-01, 1.2451e-02, + 8.5449e-03, 1.2109e-01, 5.5847e-03, 5.2344e-01, 3.0762e-02, + 5.0293e-02, 1.6479e-02, 3.8086e-01, 7.7209e-03, 1.8433e-02, + 8.2520e-02, 1.2817e-02, 1.8799e-02, 4.6289e-01, 2.1057e-03, + 2.6978e-02, 5.4688e-02, 8.0078e-02, 6.3171e-03, 1.4343e-02, + 4.7913e-03, 2.3804e-02, 1.0681e-02, 3.6133e-01, 1.6479e-02, + 3.1281e-03, 2.0630e-02, 5.9570e-02, 5.9814e-02, 1.3611e-02, + 7.0496e-03, 6.0156e-01, 1.6022e-04, 5.7031e-01, 1.4038e-02, + -4.2114e-03, 1.6602e-02, 3.0664e-01, 2.7344e-02, 3.4637e-03, + 1.1035e-01, 5.0964e-03, 6.2988e-02, 8.8501e-03, 1.9409e-02, + 8.2397e-03, 1.3000e-02, 6.5918e-02, 2.0508e-01, 1.0193e-02, + 1.7548e-03, 1.7212e-02, 4.6631e-02, 2.4170e-02, 2.6978e-02, + 1.5869e-02, 1.7334e-02, 2.3193e-02, 6.3965e-02, 2.7344e-02, + 4.2725e-02, 1.5320e-02, 3.8452e-03, 6.5918e-02, 5.8594e-01, + 7.6172e-02, 6.6223e-03, 4.8047e-01, 1.3062e-02, 3.0518e-02, + 1.8066e-02, 1.2939e-02, 1.5259e-04, 9.0332e-02, 4.7852e-01, + 2.3682e-02, 2.4292e-02, 5.7422e-01, 1.2305e-01, 2.6245e-02, + 1.9653e-02, 6.9141e-01, 1.5137e-01, 1.4648e-02, 1.3916e-02, + 5.4321e-03, 2.7954e-02, 1.1108e-02], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.v_proj.weight': tensor([[ 0.0136, -0.0051, -0.0161, ..., 0.0037, -0.0212, -0.0067], + [-0.0162, -0.0009, -0.0035, ..., -0.0041, -0.0212, 0.0166], + [-0.0165, -0.0024, 0.0033, ..., 0.0016, 0.0272, 0.0325], + ..., + [ 0.0039, -0.0062, -0.0161, ..., 0.0018, 0.0115, 0.0039], + [ 0.0231, 0.0065, 0.0222, ..., 0.0047, 0.0070, 0.0078], + [ 0.0183, 0.0043, 0.0031, ..., 0.0067, -0.0141, -0.0238]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.output_proj.weight': tensor([[ 0.0136, 0.0078, -0.0025, ..., -0.0070, 0.0131, 0.0087], + [-0.0044, -0.0047, 0.0051, ..., 0.0121, 0.0042, -0.0007], + [ 0.0028, -0.0032, 0.0048, ..., -0.0025, 0.0024, -0.0036], + ..., + [-0.0141, -0.0090, -0.0069, ..., 0.0048, -0.0078, 0.0093], + [ 0.0165, -0.0115, 0.0109, ..., -0.0265, -0.0118, -0.0236], + [ 0.0045, -0.0073, 0.0156, ..., 0.0239, 0.0071, 0.0122]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0066, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [ 0.0018, 0.0150, -0.0159, ..., 0.0040, -0.0058, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0070, -0.0072], + [-0.0337, -0.0037, 0.0052, ..., -0.0039, 0.0172, 0.0231], + [-0.0255, -0.0015, -0.0039, ..., 0.0408, 0.0430, -0.0123]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0031], + [ 0.0198, -0.0208, 0.0129, ..., 0.0088, -0.0337, -0.0051], + [ 0.0302, 0.0101, 0.0025, ..., -0.0304, 0.0199, -0.0019], + ..., + [ 0.0171, -0.0276, -0.0075, ..., 0.0231, 0.0182, -0.0045], + [ 0.0221, 0.0388, 0.0060, ..., 0.0037, -0.0488, -0.0361], + [ 0.0040, 0.0115, -0.0058, ..., 0.0037, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img.weight': tensor([[ 0.0273, 0.0038, 0.0118, ..., -0.0015, 0.0347, 0.0615], + [-0.0258, 0.0009, -0.0197, ..., 0.0139, -0.0045, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0096, 0.0132, 0.0184, ..., 0.0110, 0.0084, 0.0515], + [-0.0010, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.mlp.layer1.weight': tensor([[-0.0015, 0.0028, 0.0081, ..., 0.0044, -0.0105, -0.0049], + [-0.0078, 0.0064, 0.0078, ..., -0.0059, -0.0128, 0.0036], + [-0.0074, 0.0156, 0.0066, ..., 0.0040, -0.0019, -0.0508], + ..., + [-0.0068, 0.0077, -0.0031, ..., 0.0070, 0.0021, -0.0077], + [ 0.0013, -0.0021, -0.0081, ..., -0.0139, -0.0040, -0.0269], + [ 0.0014, 0.0190, 0.0085, ..., 0.0118, 0.0143, -0.0080]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.mlp.layer2.weight': tensor([[-0.0039, -0.0052, -0.0156, ..., -0.0110, -0.0114, 0.0038], + [ 0.0078, -0.0022, 0.0026, ..., -0.0114, -0.0106, 0.0042], + [-0.0045, 0.0028, -0.0003, ..., 0.0067, 0.0045, 0.0061], + ..., + [-0.0017, -0.0100, -0.0094, ..., 0.0044, 0.0038, 0.0044], + [-0.0137, -0.0057, 0.0025, ..., 0.0140, 0.0033, -0.0125], + [ 0.0079, 0.0063, -0.0026, ..., -0.0060, 0.0133, -0.0347]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0299, 0.0299, 0.0286, ..., -0.0041, -0.0042, -0.0043], + [-0.0284, -0.0275, -0.0271, ..., 0.0042, 0.0046, 0.0045], + [-0.0298, -0.0292, -0.0287, ..., 0.0044, 0.0048, 0.0048], + ..., + [-0.0344, -0.0332, -0.0327, ..., 0.0020, 0.0021, 0.0023], + [-0.0337, -0.0332, -0.0325, ..., 0.0020, 0.0021, 0.0020], + [-0.0156, -0.0157, -0.0159, ..., 0.0001, 0.0001, 0.0001]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0055, -0.0037, -0.0053, ..., -0.0017, -0.0019, -0.0014], + [ 0.0044, -0.0040, -0.0042, ..., -0.0033, -0.0038, -0.0022], + [ 0.0057, -0.0066, -0.0056, ..., -0.0018, -0.0021, 0.0015], + ..., + [ 0.0420, -0.0452, -0.0454, ..., -0.0278, -0.0244, 0.0075], + [-0.0060, 0.0142, 0.0075, ..., 0.0167, 0.0153, 0.0170], + [-0.0189, 0.0197, 0.0195, ..., 0.0182, 0.0177, 0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5420e-02, 5.4443e-02, 5.3467e-02, ..., -1.3733e-04, + -6.7234e-05, -2.1935e-04], + [ 5.4688e-02, 5.2979e-02, 5.1025e-02, ..., 2.7275e-04, + 3.8528e-04, 4.1771e-04], + [-4.5654e-02, -4.4434e-02, -4.3213e-02, ..., -3.8445e-06, + 5.1498e-05, -5.6744e-05], + ..., + [ 2.8809e-02, 2.7954e-02, 2.7466e-02, ..., 1.0109e-04, + 3.3140e-05, 6.2943e-05], + [ 5.9814e-02, 5.8594e-02, 5.7861e-02, ..., 4.6921e-04, + 5.1117e-04, 5.0354e-04], + [-4.7607e-02, -4.6631e-02, -4.5654e-02, ..., -2.9755e-04, + -3.4142e-04, -3.0327e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0041, 0.0135, -0.0018, ..., -0.0096, -0.0148, 0.0106], + [-0.0134, 0.0061, 0.0108, ..., -0.0055, -0.0068, 0.0040], + [-0.0006, 0.0032, 0.0011, ..., -0.0012, -0.0017, 0.0017], + ..., + [ 0.0178, 0.0061, -0.0118, ..., 0.0077, 0.0155, 0.0041], + [-0.0043, 0.0393, 0.0046, ..., -0.0013, 0.0275, -0.0136], + [-0.0101, 0.0148, 0.0110, ..., -0.0106, -0.0070, 0.0193]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_mlp.1.weight': tensor([[-0.0518, -0.0491, -0.0476, ..., 0.0030, 0.0032, 0.0032], + [-0.0063, 0.0039, 0.0039, ..., 0.0007, 0.0008, 0.0007], + [ 0.0645, 0.0625, 0.0623, ..., -0.0016, -0.0016, -0.0016], + ..., + [ 0.0588, 0.0574, 0.0566, ..., -0.0017, -0.0017, -0.0015], + [ 0.0151, 0.0139, 0.0133, ..., -0.0010, -0.0009, -0.0010], + [-0.0598, -0.0586, -0.0576, ..., 0.0021, 0.0021, 0.0020]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_mlp.2.weight': tensor([[-0.0337, -0.0108, 0.0143, ..., 0.0069, 0.0125, -0.0147], + [ 0.0018, -0.0047, -0.0014, ..., -0.0011, 0.0047, 0.0011], + [ 0.0029, -0.0035, -0.0018, ..., -0.0018, 0.0041, 0.0018], + ..., + [ 0.0403, -0.0041, -0.0137, ..., -0.0083, 0.0041, 0.0143], + [ 0.0270, 0.0325, 0.0256, ..., 0.0282, -0.0325, -0.0258], + [ 0.0095, -0.0148, -0.0086, ..., -0.0112, 0.0151, 0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.after_proj.weight': tensor([[-0.0071, -0.0042, 0.0067, ..., -0.0049, 0.0068, -0.0044], + [ 0.0003, 0.0037, 0.0046, ..., -0.0112, -0.0070, -0.0019], + [-0.0046, -0.0081, 0.0173, ..., 0.0039, 0.0036, 0.0035], + ..., + [-0.0026, 0.0045, 0.0090, ..., -0.0181, 0.0084, 0.0073], + [ 0.0036, -0.0075, -0.0081, ..., 0.0049, -0.0157, 0.0065], + [-0.0063, -0.0060, 0.0150, ..., -0.0077, 0.0002, 0.0079]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.after_proj.bias': tensor([-0.0017, -0.0015, 0.0014, ..., -0.0042, 0.0008, 0.0029], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_proj.weight': tensor([[-0.0034, 0.0080, 0.0164, ..., -0.0070, 0.0063, -0.0270], + [-0.0066, 0.0023, -0.0067, ..., -0.0078, -0.0089, -0.0082], + [ 0.0114, 0.0062, 0.0042, ..., -0.0023, -0.0038, -0.0038], + ..., + [ 0.0045, -0.0076, -0.0040, ..., -0.0042, 0.0046, 0.0069], + [-0.0044, -0.0320, -0.0098, ..., -0.0381, -0.0189, -0.0298], + [-0.0075, -0.0038, -0.0040, ..., 0.0037, -0.0047, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_norm.weight': tensor([ 1.2656e+00, 1.6562e+00, 1.5078e+00, 1.7188e+00, 1.0498e-01, + 6.9824e-02, 4.7607e-02, 1.0192e-05, -1.2757e-30, -1.5564e-03, + 2.6978e-02, 1.0156e-01, -2.2095e-02, -2.9297e-02, 9.0790e-04, + 3.2616e-04, 6.0059e-02, 9.4922e-01, 9.9609e-01, 1.4267e-03, + 6.1035e-05, 1.1953e+00, 1.6406e+00, 1.6719e+00, 6.7969e-01, + 6.2891e-01, 1.5703e+00, -3.4332e-03, -1.0071e-02, 2.0146e-05, + 4.3869e-04, -5.0354e-03, 2.3682e-02, 1.0938e+00, 5.4932e-03, + 1.0781e+00, 1.5068e-04, 4.7207e-05, 1.3123e-02, -1.6327e-03, + 8.5449e-03, -3.2187e-05, 1.2054e-03, 1.6953e+00, 9.8438e-01, + 1.4062e+00, 1.6406e+00, 1.8516e+00, 1.0596e-01, 4.9973e-04, + 3.0029e-02, 2.2168e-01, 1.0859e+00, 8.2520e-02, 2.1820e-03, + 1.3086e-01, 5.6458e-04, 9.3750e-02, 7.4863e-05, -1.1902e-02, + 1.4267e-03, 3.8300e-03, -3.3722e-03, 1.1406e+00, 1.8828e+00, + 7.9297e-01, 9.5703e-01, 1.5312e+00, 4.0771e-02, -1.2390e-02, + 1.2085e-02, -1.9908e-05, 2.5761e-30, 1.1658e-02, 1.7822e-02, + 8.9722e-03, 6.0730e-03, 4.0588e-03, -2.5269e-02, 1.8066e-02, + 5.7861e-02, -2.6245e-03, 1.4587e-02, 2.7618e-03, 3.3379e-04, + 4.1504e-03, 5.5469e-01, 4.2578e-01, 1.4688e+00, 1.6406e+00, + 4.7656e-01, 5.7068e-03, -7.9727e-04, 2.9755e-04, -4.3335e-03, + 1.0300e-03, 3.4912e-02, 6.8848e-02, 5.2795e-03, 5.2002e-02, + 2.6321e-04, 4.1485e-05, 1.2891e-01, -2.6321e-04, 6.5918e-03, + -1.4484e-05, 1.3638e-04, 4.1797e-01, 1.5000e+00, 1.4062e+00, + 5.9375e-01, 9.2969e-01, 7.3242e-02, -2.4109e-03, 2.0020e-02, + 1.1641e+00, 5.0049e-02, 1.0859e+00, -1.5736e-05, 2.3651e-03, + 3.9307e-02, 4.5410e-02, 1.4210e-04, 1.9302e-03, -1.0872e-04, + -3.6011e-03, 9.4141e-01, -2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.self_attn.k_proj.weight': tensor([[-4.2725e-02, 1.7700e-03, -2.9449e-03, ..., -3.2959e-03, + -3.9101e-05, -1.3123e-02], + [-3.5156e-02, -1.2817e-02, -1.3123e-02, ..., 1.0071e-02, + -1.1230e-02, 2.4414e-02], + [ 4.5410e-02, 4.3030e-03, 3.8910e-03, ..., -1.2146e-02, + -4.1809e-03, -2.7100e-02], + ..., + [-7.1411e-03, 1.5259e-02, 5.7602e-04, ..., -5.2795e-03, + -6.2561e-03, -2.8076e-03], + [ 1.1597e-02, -8.8501e-03, -2.4261e-03, ..., -4.0894e-03, + 1.0986e-02, 3.2349e-03], + [ 2.0294e-03, 5.9814e-03, -1.2970e-03, ..., -5.9509e-03, + -6.1035e-04, 6.5308e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.k_norm.weight': tensor([ 7.4609e-01, 1.5938e+00, 1.5234e+00, 1.6562e+00, 1.0742e-01, + 7.5195e-02, 4.6387e-02, -1.3113e-05, 8.1351e-31, -2.1210e-03, + 2.6611e-02, 1.0205e-01, -2.6245e-02, 5.2979e-02, 1.1597e-03, + 1.4114e-04, 4.0588e-03, 9.4922e-01, 9.9609e-01, 1.3275e-03, + 1.0681e-04, 1.1953e+00, 1.4531e+00, 1.5469e+00, 3.5400e-02, + 1.4355e-01, 1.5938e+00, -5.7373e-03, 9.6436e-03, -9.6798e-05, + -1.5640e-03, 5.0964e-03, -2.0630e-02, 1.0938e+00, -2.0142e-03, + 1.0703e+00, 1.6975e-04, 4.4584e-05, -1.5076e-02, -1.7395e-03, + 8.5449e-03, -7.4387e-05, 1.2054e-03, 1.4609e+00, 4.4727e-01, + 1.3750e+00, 1.6953e+00, 1.8672e+00, -5.3955e-02, 1.8387e-03, + 1.0681e-02, 1.5625e-02, 1.0859e+00, 7.6172e-02, 2.1820e-03, + 1.3086e-01, -9.0122e-05, 9.3750e-02, 7.0095e-05, 7.6660e-02, + -1.4191e-03, 1.8311e-02, -1.9653e-02, 1.1406e+00, 1.9688e+00, + 1.1172e+00, 8.6328e-01, 1.5781e+00, 2.7710e-02, -5.7983e-03, + -1.3916e-02, 1.5974e-05, 1.9352e-30, 1.2024e-02, 3.1006e-02, + 5.0781e-02, 6.3782e-03, 2.6611e-02, -3.7842e-02, -1.8066e-02, + -1.8555e-02, 6.0730e-03, 9.0332e-03, 2.6550e-03, 3.2425e-04, + 8.6670e-03, 1.1172e+00, 1.0000e+00, 1.5625e+00, 1.7031e+00, + 1.4844e-01, 5.7678e-03, -3.0212e-03, -2.6131e-04, -3.6163e-03, + -6.2943e-04, 1.9287e-02, 9.0332e-03, 4.6387e-03, -8.9722e-03, + 2.7084e-04, -1.6332e-05, 1.2793e-01, -2.6321e-04, 6.5613e-03, + 2.1100e-05, 1.3256e-04, 1.1719e+00, 1.6406e+00, 1.4141e+00, + 1.1768e-01, 9.8438e-01, 8.7280e-03, 4.0894e-03, 5.0354e-03, + 1.1641e+00, -2.9755e-03, 1.0938e+00, -7.2002e-05, -1.8158e-03, + 3.9307e-02, 4.5166e-02, 1.3924e-04, 4.3457e-02, -1.1730e-04, + 1.5015e-02, 9.4141e-01, -1.5503e-02], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.self_attn.v_proj.weight': tensor([[-0.0017, 0.0042, 0.0115, ..., 0.0542, -0.0092, 0.0073], + [ 0.0060, 0.0214, -0.0223, ..., -0.0248, -0.0010, -0.0054], + [-0.0009, 0.0156, -0.0039, ..., 0.0126, -0.0044, 0.0096], + ..., + [ 0.0026, -0.0039, -0.0148, ..., 0.0102, 0.0055, -0.0009], + [ 0.0059, 0.0243, 0.0019, ..., 0.0025, 0.0025, 0.0002], + [-0.0074, -0.0142, 0.0027, ..., 0.0175, -0.0077, -0.0128]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.output_proj.weight': tensor([[ 0.0179, 0.0205, -0.0031, ..., 0.0021, 0.0260, -0.0159], + [ 0.0096, -0.0099, -0.0095, ..., -0.0045, -0.0166, 0.0025], + [ 0.0078, 0.0236, -0.0091, ..., 0.0062, -0.0017, -0.0050], + ..., + [-0.0287, -0.0366, -0.0045, ..., -0.0045, -0.0148, 0.0040], + [ 0.0026, 0.0087, -0.0182, ..., -0.0103, 0.0327, -0.0085], + [-0.0432, -0.0049, 0.0156, ..., 0.0238, -0.0083, -0.0036]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_proj.weight': tensor([[ 0.0049, 0.0067, 0.0038, ..., -0.0072, -0.0022, 0.0223], + [ 0.0061, -0.0040, -0.0039, ..., -0.0063, 0.0070, 0.0106], + [ 0.0023, 0.0029, -0.0030, ..., 0.0027, 0.0157, 0.0051], + ..., + [ 0.0076, -0.0088, -0.0059, ..., 0.0037, -0.0072, -0.0011], + [ 0.0017, -0.0032, -0.0029, ..., 0.0056, -0.0051, -0.0022], + [-0.0040, -0.0054, 0.0039, ..., -0.0080, 0.0072, 0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_norm.weight': tensor([0.1089, 0.0232, 0.4688, 0.0752, 0.1816, 0.0356, 0.1494, 0.1260, 0.5234, + 0.0388, 0.0630, 0.0148, 0.0630, 0.0762, 0.0508, 0.2021, 0.4883, 0.0260, + 0.7383, 0.0410, 0.0311, 0.1641, 0.0289, 0.0806, 0.0557, 0.4199, 0.5156, + 0.2754, 0.0713, 0.0454, 0.0233, 0.2451, 0.7031, 0.2617, 0.0415, 0.0344, + 0.0190, 0.1025, 0.0284, 0.0422, 0.0586, 0.1318, 0.0247, 0.0344, 0.2246, + 0.0425, 0.4043, 0.0304, 0.0540, 0.0095, 0.0259, 0.4141, 0.4902, 0.3867, + 0.0962, 0.6133, 0.5781, 0.0337, 0.1494, 0.0249, 0.0898, 0.4902, 0.0215, + 0.5547, 0.1699, 0.4180, 0.1040, 0.0488, 0.0723, 0.4941, 0.0262, 0.4902, + 0.0198, 0.0447, 0.1436, 0.0850, 0.0280, 0.1128, 0.6445, 0.0308, 0.0908, + 0.0874, 0.0417, 0.0381, 0.4297, 0.0723, 0.0515, 0.0874, 0.0211, 0.1787, + 0.0398, 0.1924, 0.0403, 0.0669, 0.0442, 0.1133, 0.0052, 0.3691, 0.5234, + 0.0471, 0.0581, 0.3301, 0.0347, 0.1270, 0.0352, 0.2656, 0.0214, 0.5664, + 0.5664, 0.0391, 0.0176, 0.7500, 0.0310, 0.0139, 0.5039, 0.2217, 0.3145, + 0.7109, 0.0332, 0.0913, 0.0884, 0.5391, 0.6016, 0.0146, 0.0248, 0.2041, + 0.0415, 0.0378], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_proj.weight': tensor([[-1.7548e-03, 1.8158e-03, 3.5706e-03, ..., -3.2349e-03, + 1.1353e-02, -1.2512e-02], + [ 5.9204e-03, -6.7711e-05, -2.9602e-03, ..., -3.0518e-04, + -1.3275e-03, 6.3171e-03], + [-1.1292e-02, 7.6904e-03, -1.3245e-02, ..., 6.8970e-03, + 9.8267e-03, 1.7853e-03], + ..., + [-1.6479e-03, 3.1128e-03, -5.8899e-03, ..., -3.5095e-03, + -1.3306e-02, 9.3384e-03], + [-1.2451e-02, -1.8539e-03, -8.9111e-03, ..., 7.5684e-03, + -9.2163e-03, -2.7275e-04], + [ 1.4893e-02, -1.4038e-03, 5.0049e-03, ..., -1.0071e-02, + 4.8523e-03, -3.5400e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_norm.weight': tensor([0.1084, 0.0232, 0.4688, 0.0752, 0.1816, 0.0361, 0.1494, 0.1260, 0.5234, + 0.0393, 0.0640, 0.0149, 0.0630, 0.0762, 0.0510, 0.2041, 0.4883, 0.0262, + 0.7344, 0.0410, 0.0309, 0.1641, 0.0291, 0.0801, 0.0557, 0.4160, 0.5156, + 0.2734, 0.0713, 0.0452, 0.0242, 0.2471, 0.7031, 0.2617, 0.0422, 0.0344, + 0.0188, 0.1025, 0.0287, 0.0417, 0.0583, 0.1328, 0.0251, 0.0349, 0.2246, + 0.0425, 0.4043, 0.0306, 0.0542, 0.0092, 0.0262, 0.4141, 0.4902, 0.3867, + 0.0962, 0.6133, 0.5820, 0.0334, 0.1494, 0.0253, 0.0898, 0.4902, 0.0219, + 0.5547, 0.1689, 0.4180, 0.1040, 0.0488, 0.0723, 0.4961, 0.0261, 0.4902, + 0.0195, 0.0444, 0.1436, 0.0845, 0.0280, 0.1138, 0.6445, 0.0305, 0.0898, + 0.0874, 0.0417, 0.0383, 0.4297, 0.0728, 0.0505, 0.0869, 0.0214, 0.1807, + 0.0408, 0.1914, 0.0400, 0.0674, 0.0447, 0.1143, 0.0052, 0.3691, 0.5234, + 0.0464, 0.0588, 0.3301, 0.0347, 0.1270, 0.0352, 0.2656, 0.0211, 0.5664, + 0.5664, 0.0391, 0.0177, 0.7500, 0.0305, 0.0141, 0.5078, 0.2217, 0.3145, + 0.7070, 0.0325, 0.0898, 0.0884, 0.5391, 0.6016, 0.0145, 0.0248, 0.2041, + 0.0427, 0.0381], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.v_proj.weight': tensor([[ 0.0099, -0.0024, -0.0236, ..., -0.0018, -0.0025, -0.0165], + [-0.0064, 0.0107, 0.0040, ..., 0.0040, 0.0023, -0.0250], + [ 0.0190, 0.0034, -0.0056, ..., 0.0033, -0.0052, -0.0396], + ..., + [ 0.0025, 0.0031, 0.0190, ..., -0.0087, -0.0259, 0.0049], + [-0.0332, -0.0007, -0.0219, ..., -0.0072, 0.0103, -0.0148], + [-0.0070, -0.0017, -0.0017, ..., -0.0082, 0.0104, 0.0071]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.output_proj.weight': tensor([[-0.0128, -0.0076, -0.0082, ..., -0.0063, 0.0133, 0.0111], + [-0.0248, -0.0042, -0.0100, ..., 0.0048, -0.0138, -0.0304], + [ 0.0011, 0.0024, 0.0064, ..., 0.0029, -0.0026, -0.0045], + ..., + [ 0.0141, -0.0109, 0.0123, ..., 0.0160, -0.0127, -0.0093], + [-0.0161, -0.0079, 0.0060, ..., 0.0239, -0.0284, 0.0334], + [-0.0165, -0.0115, 0.0254, ..., -0.0051, 0.0259, -0.0051]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_img.weight': tensor([[ 1.6235e-02, -4.2480e-02, -2.7954e-02, ..., 2.3682e-02, + -4.1748e-02, -4.6631e-02], + [-1.0925e-02, 2.9297e-02, 2.5635e-02, ..., -2.3315e-02, + -1.9775e-02, -2.3193e-02], + [-1.2146e-02, 4.9438e-03, -6.5918e-03, ..., 2.6611e-02, + -2.7832e-02, 7.5989e-03], + ..., + [-2.1729e-02, -4.1016e-02, 4.8828e-02, ..., 3.8862e-05, + -2.3956e-03, -2.5513e-02], + [ 9.9182e-04, 6.6833e-03, 3.8300e-03, ..., 3.0029e-02, + 8.4229e-03, 2.0508e-02], + [-2.1362e-02, 2.1729e-02, 5.6458e-03, ..., -1.3916e-02, + -1.4282e-02, -4.1992e-02]], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, 0.0013, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0024, 0.0293, ..., 0.0067, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0139, 0.0126, -0.0092], + ..., + [-0.0032, 0.0320, -0.0233, ..., -0.0062, -0.0028, 0.0188], + [ 0.0135, -0.0352, 0.0267, ..., -0.0305, -0.0039, 0.0018], + [-0.0205, 0.0140, 0.0039, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, 0.0005, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0056, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0015], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, 0.0013, 0.0012], + [-0.0141, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0023, ..., -0.0103, 0.0071, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.mlp.layer1.weight': tensor([[ 0.0104, 0.0049, -0.0160, ..., 0.0096, 0.0089, -0.0065], + [-0.0109, -0.0065, 0.0049, ..., -0.0120, -0.0062, -0.0079], + [-0.0034, 0.0359, -0.0165, ..., 0.0284, 0.0075, 0.0089], + ..., + [-0.0074, 0.0103, 0.0645, ..., -0.0041, 0.0085, -0.0117], + [-0.0038, 0.0049, 0.0042, ..., -0.0151, 0.0087, -0.0129], + [ 0.0055, 0.0066, -0.0072, ..., -0.0146, -0.0156, -0.0186]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.mlp.layer2.weight': tensor([[ 0.0111, -0.0063, 0.0005, ..., -0.0266, -0.0067, 0.0039], + [-0.0112, 0.0051, -0.0106, ..., -0.0027, 0.0101, -0.0044], + [ 0.0006, -0.0107, 0.0055, ..., 0.0058, 0.0098, -0.0042], + ..., + [ 0.0040, 0.0040, 0.0217, ..., -0.0044, 0.0064, -0.0055], + [ 0.0053, 0.0029, 0.0039, ..., 0.0110, 0.0036, 0.0095], + [-0.0168, -0.0084, 0.0029, ..., 0.0116, 0.0139, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0413, -0.0408, -0.0405, ..., 0.0009, 0.0007, 0.0008], + [-0.0369, -0.0366, -0.0356, ..., 0.0026, 0.0027, 0.0027], + [-0.0334, -0.0332, -0.0320, ..., 0.0021, 0.0022, 0.0022], + ..., + [ 0.0244, 0.0234, 0.0225, ..., -0.0037, -0.0038, -0.0037], + [ 0.0134, 0.0123, 0.0115, ..., -0.0032, -0.0031, -0.0031], + [-0.0344, -0.0337, -0.0330, ..., 0.0022, 0.0023, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0031, 0.0044, ..., 0.0063, 0.0090, 0.0045], + [-0.0028, -0.0012, -0.0015, ..., -0.0019, -0.0014, -0.0011], + [ 0.0024, -0.0069, -0.0049, ..., 0.0032, 0.0033, -0.0058], + ..., + [-0.0237, -0.0811, -0.0635, ..., 0.0327, 0.0332, -0.0723], + [-0.0161, -0.0140, -0.0139, ..., -0.0115, -0.0139, -0.0131], + [ 0.0098, 0.0067, 0.0069, ..., 0.0092, 0.0121, 0.0064]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_cross_attn.1.weight': tensor([[-7.9102e-02, -7.6660e-02, -7.5195e-02, ..., -9.5844e-05, + -1.1110e-04, -1.0157e-04], + [ 3.0273e-02, 2.9785e-02, 2.9297e-02, ..., 2.5481e-06, + 2.8610e-05, 1.4877e-04], + [ 2.5391e-02, 2.4658e-02, 2.3682e-02, ..., 2.4796e-04, + 2.7084e-04, 1.7738e-04], + ..., + [ 5.1514e-02, 5.0293e-02, 4.9805e-02, ..., -1.8692e-04, + -1.8120e-04, -1.1349e-04], + [ 5.5176e-02, 5.3711e-02, 5.2246e-02, ..., -1.8311e-04, + -1.2207e-04, -1.7071e-04], + [-1.6235e-02, -1.5747e-02, -1.5503e-02, ..., -5.0306e-05, + 1.0109e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0053, -0.0033, ..., 0.0281, 0.0317, 0.0069], + [ 0.0040, -0.0044, 0.0056, ..., -0.0009, 0.0025, -0.0019], + [ 0.0025, -0.0038, 0.0039, ..., -0.0026, -0.0026, -0.0044], + ..., + [-0.0069, -0.0052, 0.0403, ..., 0.0016, 0.0123, -0.0041], + [-0.0302, 0.0184, 0.0298, ..., 0.0115, 0.0130, -0.0046], + [ 0.0261, -0.0206, 0.0040, ..., 0.0048, 0.0107, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_mlp.1.weight': tensor([[-7.6172e-02, -7.3730e-02, -7.2754e-02, ..., 8.7261e-05, + 2.0409e-04, 1.5354e-04], + [ 5.8105e-02, 5.5420e-02, 5.5176e-02, ..., 3.8719e-04, + 4.6921e-04, 4.5776e-04], + [-9.2773e-02, -9.1309e-02, -8.9844e-02, ..., -4.7874e-04, + -4.1199e-04, -4.4250e-04], + ..., + [-7.4707e-02, -7.2754e-02, -7.1777e-02, ..., 1.1444e-04, + 2.1648e-04, 1.1492e-04], + [-6.0791e-02, -5.9326e-02, -5.8594e-02, ..., 1.5736e-04, + 1.8477e-05, 2.9922e-05], + [ 7.2266e-02, 7.1289e-02, 7.0312e-02, ..., -9.0599e-05, + -1.1873e-04, -8.2016e-05]], dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_mlp.2.weight': tensor([[-4.9438e-03, -3.9978e-03, 2.5177e-03, ..., -7.4158e-03, + -1.1414e-02, 8.9111e-03], + [-3.3417e-03, -2.6550e-03, -1.6861e-03, ..., -2.7618e-03, + -2.5787e-03, 2.5787e-03], + [-6.4850e-05, -2.4414e-03, -1.7471e-03, ..., -3.5477e-04, + -1.5259e-04, 1.3161e-04], + ..., + [-4.6143e-02, -6.9275e-03, -5.6396e-02, ..., -3.4424e-02, + -1.2695e-02, 2.6489e-02], + [ 8.1177e-03, -2.7344e-02, 3.1738e-02, ..., 6.6528e-03, + 5.4016e-03, -6.5613e-03], + [ 9.4604e-03, 7.5989e-03, 4.9438e-03, ..., 8.5449e-03, + 4.1199e-03, -6.7444e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.after_proj.weight': tensor([[ 0.0049, 0.0079, -0.0066, ..., 0.0017, 0.0227, 0.0034], + [-0.0020, -0.0085, -0.0080, ..., 0.0134, -0.0012, -0.0102], + [-0.0118, 0.0030, -0.0039, ..., -0.0023, -0.0021, 0.0011], + ..., + [-0.0079, -0.0026, 0.0031, ..., -0.0070, 0.0021, -0.0082], + [-0.0040, 0.0076, 0.0044, ..., 0.0051, -0.0108, -0.0024], + [ 0.0119, 0.0110, 0.0015, ..., -0.0018, 0.0058, 0.0067]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.after_proj.bias': tensor([ 0.0050, -0.0118, -0.0096, ..., -0.0070, 0.0018, 0.0004], + dtype=torch.bfloat16)} +[06-26 07:35:42|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:42|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 07:35:42|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 1 checkpoint. +[06-26 07:35:42|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/depth(626e6618-bfcd-4d9a-a077-1409e2ce353f) +[06-26 07:35:42|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 2 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/dea7737ca29dd8d9086413c6dc5724b8250a0bb4/general/depth/626e6618-bfcd-4d9a-a077-1409e2ce353f_ema_bf16.pt +[06-26 07:35:43|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0144, -0.0021, 0.0064], + [ 0.0036, 0.0057, -0.0327, ..., -0.0082, -0.0024, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0137, -0.0029, 0.0283], + ..., + [ 0.0010, 0.0125, 0.0248, ..., 0.0437, -0.0074, -0.0192], + [ 0.0277, 0.0146, 0.0036, ..., -0.0031, -0.0092, 0.0146], + [ 0.0437, 0.0366, 0.0085, ..., 0.0013, -0.0128, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0028, -0.0310, 0.0300, ..., -0.0071, -0.0042, -0.0339], + [-0.0103, -0.0048, 0.0115, ..., 0.0049, 0.0092, 0.0520], + [-0.0064, 0.0132, 0.0398, ..., -0.0097, -0.0179, -0.0070], + ..., + [-0.0046, 0.0175, -0.0094, ..., 0.0225, 0.0047, -0.0020], + [-0.0209, -0.0081, 0.0087, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0354, -0.0082, ..., 0.0123, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0016, ..., 0.0376, 0.0532, 0.0011], + [ 0.0175, -0.0347, 0.0015, ..., -0.0064, 0.0047, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0156, 0.0262, -0.0114], + [-0.0015, 0.0244, 0.0356, ..., 0.0018, 0.0170, 0.0112], + [ 0.0486, 0.0175, 0.0042, ..., 0.0046, -0.0021, 0.0008]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0118, ..., -0.0075, -0.0337, 0.0079], + [ 0.0491, 0.0413, 0.0251, ..., -0.0041, 0.0393, 0.0035], + [-0.0162, -0.0005, 0.0134, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0014, 0.0261, 0.0042, ..., -0.0038, 0.0347, 0.0036], + [-0.0249, 0.0067, 0.0024, ..., 0.0096, 0.0388, -0.0205], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0123, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0048, -0.0270, -0.0166, ..., 0.0118, -0.0298, 0.0532], + [ 0.0034, 0.0036, -0.0126, ..., -0.0161, 0.0152, 0.0308], + [ 0.0145, 0.0086, 0.0137, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0067, -0.0630, 0.0027, ..., 0.0072, -0.0183, 0.0217], + [ 0.0039, 0.0032, 0.0258, ..., -0.0486, -0.0349, -0.0039], + [ 0.0182, -0.0081, 0.0197, ..., 0.0151, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0053, -0.0042, -0.0005, ..., -0.0007, -0.0017, -0.0121], + [-0.0152, 0.0251, 0.0432, ..., -0.0051, 0.0039, -0.0038], + [-0.0149, -0.0203, -0.0245, ..., -0.0118, -0.0325, 0.0029], + ..., + [ 0.0043, 0.0008, 0.0417, ..., 0.0280, 0.0022, -0.0024], + [-0.0039, 0.0037, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0067, -0.0034, -0.0085]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0155, -0.0063, -0.0481, ..., 0.0449, 0.0212, 0.0128], + [ 0.0156, -0.0227, 0.0220, ..., -0.0219, 0.0005, 0.0081], + [-0.0156, -0.0098, 0.0430, ..., -0.0101, 0.0327, -0.0110], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0136], + [ 0.0021, 0.0166, -0.0061, ..., 0.0013, 0.0079, -0.0052], + [-0.0110, -0.0041, 0.0117, ..., 0.0047, -0.0034, -0.0077]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0035, -0.0386, -0.0120, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0075, 0.0286, ..., -0.0339, -0.0160, 0.0170], + [-0.0127, -0.0053, -0.0128, ..., 0.0214, 0.0111, -0.0044], + ..., + [-0.0153, -0.0172, -0.0137, ..., 0.0098, 0.0098, 0.0184], + [-0.0154, -0.0356, 0.0201, ..., -0.0069, -0.0117, 0.0332], + [-0.0396, -0.0121, -0.0090, ..., -0.0284, -0.0141, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0280, 0.0312, 0.0133, ..., -0.0195, 0.0114, -0.0153], + [ 0.0192, 0.0092, -0.0073, ..., 0.0022, -0.0127, -0.0198], + [-0.0249, 0.0109, -0.0137, ..., 0.0090, 0.0172, 0.0067], + ..., + [ 0.0466, -0.0045, 0.0007, ..., 0.0010, 0.0239, 0.0042], + [ 0.0547, -0.0194, -0.0108, ..., -0.0069, 0.0311, -0.0044], + [ 0.0051, 0.0104, -0.0095, ..., 0.0203, 0.0170, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0044, -0.0046, 0.0283, ..., -0.0109, -0.0045, -0.0182], + [ 0.0052, 0.0126, -0.0315, ..., -0.0457, -0.0069, 0.0109], + [ 0.0006, 0.0261, 0.0245, ..., 0.0046, 0.0084, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0155, 0.0058, -0.0077], + [ 0.0059, -0.0371, 0.0356, ..., -0.0082, 0.0410, -0.0178], + [ 0.0178, -0.0155, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0087, 0.0208, 0.0272, ..., 0.0137, 0.0156, 0.0082], + [-0.0082, -0.0023, -0.0041, ..., 0.0144, 0.0007, 0.0057], + [ 0.0046, 0.0010, 0.0240, ..., -0.0119, -0.0066, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0046, ..., 0.0248, -0.0106, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., 0.0020, -0.0295, 0.0400], + [-0.0118, 0.0135, 0.0305, ..., -0.0100, 0.0366, 0.0227]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0054, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0149, -0.0054, 0.0288, ..., 0.0136, 0.0583, 0.0044], + [-0.0281, 0.0154, -0.0386, ..., -0.0444, -0.0093, -0.0009], + ..., + [ 0.0035, 0.0157, 0.0217, ..., 0.0047, 0.0115, 0.0151], + [ 0.0051, 0.0195, -0.0283, ..., -0.0048, -0.0039, -0.0342], + [-0.0167, -0.0195, 0.0007, ..., -0.0048, 0.0165, 0.0132]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0140, -0.0146, 0.0459, ..., 0.0043, -0.0320, 0.0275], + [ 0.0034, 0.0061, 0.0272, ..., -0.0219, -0.0183, -0.0265], + [ 0.0219, 0.0216, -0.0201, ..., -0.0022, 0.0052, 0.0041], + ..., + [-0.0258, 0.0028, 0.0215, ..., 0.0017, 0.0236, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0060], + [-0.0037, -0.0085, -0.0128, ..., -0.0334, 0.0062, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0146, -0.0454, ..., -0.0126, -0.0131, 0.0400], + [ 0.0295, -0.0187, -0.0106, ..., -0.0283, 0.0040, -0.0186], + [ 0.0013, 0.0547, -0.0042, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0280, 0.0045, -0.0269, ..., -0.0090, -0.0287, 0.0150], + [ 0.0093, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0098, 0.0106, ..., -0.0075, 0.0703, -0.0059]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0094, -0.0315, ..., 0.0058, 0.0254, 0.0359], + [ 0.0087, -0.0048, 0.0028, ..., -0.0125, -0.0233, -0.0042], + [ 0.0070, 0.0054, -0.0121, ..., -0.0123, 0.0018, 0.0089], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0040], + [ 0.0339, -0.0059, 0.0223, ..., 0.0410, 0.0028, 0.0138], + [-0.0055, -0.0439, -0.0486, ..., 0.0140, -0.0130, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0002, 0.0139, -0.0042, ..., -0.0079, -0.0045, -0.0229], + [-0.0289, 0.0231, 0.0084, ..., -0.0140, 0.0056, 0.0222], + [ 0.0085, 0.0295, -0.0193, ..., -0.0164, 0.0310, -0.0262], + ..., + [ 0.0233, -0.0053, -0.0029, ..., 0.0051, 0.0278, 0.0031], + [-0.0104, -0.0018, 0.0140, ..., -0.0175, 0.0444, 0.0425], + [ 0.0054, -0.0084, 0.0096, ..., 0.0061, -0.0344, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[ 0.0027, 0.0126, 0.0188, ..., -0.0092, -0.0199, 0.0147], + [-0.0581, 0.0294, -0.0043, ..., -0.0240, 0.0058, 0.0302], + [ 0.0087, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [ 0.0002, -0.0292, 0.0537, ..., 0.0033, -0.0078, -0.0106], + [-0.0036, -0.0077, 0.0209, ..., 0.0043, 0.0093, 0.0187], + [-0.0087, -0.0349, -0.0173, ..., -0.0400, 0.0057, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0182, -0.0265, -0.0105, ..., 0.0099, -0.0132, -0.0405], + [ 0.0051, -0.0087, 0.0266, ..., -0.0400, 0.0060, 0.0391], + [-0.0016, 0.0359, 0.0284, ..., 0.0292, -0.0233, 0.0212], + ..., + [ 0.0026, 0.0121, 0.0132, ..., -0.0141, -0.0048, 0.0095], + [ 0.0299, -0.0167, -0.0042, ..., -0.0013, -0.0002, -0.0206], + [ 0.0072, -0.0045, -0.0042, ..., 0.0037, -0.0022, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, -0.0015, 0.0118, ..., -0.0029, -0.0208, -0.0074], + [-0.0096, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0232], + [ 0.0134, -0.0278, 0.0118, ..., 0.0396, -0.0079, -0.0069], + ..., + [-0.0001, -0.0018, -0.0311, ..., 0.0344, 0.0051, -0.0123], + [-0.0236, -0.0102, -0.0072, ..., 0.0332, -0.0114, -0.0179], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0137, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[-0.0014, 0.0166, 0.0036, ..., 0.0273, -0.0352, 0.0156], + [-0.0510, -0.0062, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0077, ..., -0.0388, -0.0019, 0.0398], + ..., + [-0.0160, -0.0195, -0.0027, ..., 0.0037, -0.0056, 0.0029], + [-0.0187, 0.0038, 0.0098, ..., 0.0287, -0.0125, -0.0095], + [-0.0181, -0.0032, 0.0247, ..., -0.0143, 0.0052, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0096, 0.0141, ..., -0.0112, 0.0026, -0.0253], + [-0.0167, 0.0330, -0.0171, ..., -0.0037, -0.0403, -0.0110], + [ 0.0084, 0.0347, 0.0275, ..., 0.0148, 0.0064, 0.0079], + ..., + [ 0.0272, -0.0168, -0.0106, ..., -0.0098, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0059, ..., -0.0032, -0.0018, -0.0243], + [-0.0075, 0.0581, 0.0057, ..., 0.0145, 0.0239, 0.0116]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0118, -0.0410, 0.0126, ..., -0.0254, -0.0052, -0.0173], + [ 0.0476, 0.0315, 0.0148, ..., -0.0150, -0.0066, -0.0012], + [-0.0089, -0.0269, 0.0083, ..., 0.0332, 0.0144, -0.0063], + ..., + [-0.0046, -0.0043, 0.0160, ..., -0.0177, 0.0168, -0.0126], + [ 0.0051, -0.0187, -0.0049, ..., 0.0248, -0.0148, -0.0068], + [-0.0142, -0.0024, 0.0148, ..., -0.0043, 0.0037, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0054, 0.0094, ..., 0.0255, -0.0253, -0.0135], + [-0.0018, -0.0143, -0.0160, ..., -0.0113, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0115, 0.0046, 0.0322, ..., -0.0183, -0.0010, 0.0016], + [ 0.0034, -0.0050, 0.0162, ..., 0.0043, 0.0255, 0.0063], + [ 0.0270, 0.0087, 0.0159, ..., 0.0189, 0.0148, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0066, ..., -0.0532, 0.0104, 0.0036], + [-0.0240, -0.0215, -0.0103, ..., -0.0059, 0.0098, -0.0168], + [ 0.0074, -0.0013, -0.0059, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0057, 0.0172, -0.0106, ..., -0.0300, -0.0045, 0.0435], + [ 0.0051, -0.0261, 0.0247, ..., 0.0192, -0.0308, 0.0061], + [-0.0167, 0.0123, 0.0106, ..., -0.0247, 0.0123, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0078, -0.0078, 0.0317, ..., 0.0200, -0.0104, -0.0376], + [ 0.0356, 0.0378, 0.0135, ..., 0.0325, 0.0391, 0.0173], + [-0.0244, -0.0264, 0.0371, ..., 0.0028, 0.0052, -0.0322], + ..., + [ 0.0058, 0.0107, -0.0040, ..., 0.0366, -0.0206, 0.0374], + [-0.0046, -0.0070, 0.0034, ..., -0.0093, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0026, ..., -0.0166, -0.0056, 0.0245]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0053, -0.0093, -0.0344, ..., -0.0076, 0.0347, -0.0176], + [ 0.0267, -0.0366, -0.0088, ..., 0.0217, 0.0349, -0.0043], + [-0.0097, -0.0286, 0.0074, ..., 0.0052, -0.0282, 0.0013], + ..., + [-0.0129, 0.0037, 0.0058, ..., -0.0086, 0.0165, 0.0076], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0129], + [-0.0140, 0.0043, -0.0170, ..., -0.0127, 0.0060, 0.0070]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0025, -0.0047, ..., -0.0483, 0.0374, 0.0093], + [-0.0576, 0.0116, 0.0374, ..., -0.0153, -0.0188, -0.0136], + ..., + [ 0.0260, -0.0007, -0.0077, ..., -0.0003, 0.0053, -0.0165], + [-0.0222, -0.0155, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0131, -0.0148, -0.0038, ..., 0.0018, -0.0234, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0101], + [ 0.0249, 0.0205, 0.0075, ..., -0.0061, -0.0089, -0.0105], + [-0.0220, 0.0044, 0.0047, ..., 0.0309, 0.0381, 0.0037], + ..., + [ 0.0200, -0.0033, 0.0171, ..., 0.0122, 0.0320, -0.0157], + [ 0.0349, 0.0120, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0075, 0.0256, 0.0050, ..., 0.0205, 0.0035, -0.0105]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0043, 0.0199, ..., -0.0050, 0.0276, -0.0308], + [ 0.0178, -0.0063, 0.0099, ..., 0.0108, -0.0398, 0.0300], + [ 0.0430, -0.0124, 0.0081, ..., -0.0009, -0.0066, -0.0055], + ..., + [ 0.0192, -0.0106, -0.0108, ..., -0.0025, 0.0330, 0.0142], + [-0.0124, -0.0542, -0.0205, ..., 0.0062, -0.0159, 0.0320], + [ 0.0303, 0.0054, -0.0425, ..., 0.0009, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0068, -0.0315, ..., 0.0525, 0.0034, -0.0349], + [-0.0044, -0.0070, -0.0156, ..., 0.0053, -0.0240, 0.0305], + [-0.0256, -0.0045, 0.0062, ..., -0.0459, -0.0148, -0.0165], + ..., + [ 0.0070, 0.0231, 0.0209, ..., 0.0077, -0.0038, 0.0461], + [-0.0067, -0.0085, 0.0004, ..., -0.0217, 0.0347, -0.0018], + [ 0.0057, 0.0234, 0.0056, ..., 0.0261, -0.0038, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0071, 0.0023], + [-0.0183, -0.0211, 0.0045, ..., -0.0145, 0.0143, -0.0166], + [-0.0075, -0.0160, 0.0053, ..., -0.0203, 0.0267, 0.0217], + ..., + [ 0.0308, 0.0096, 0.0075, ..., 0.0171, 0.0053, -0.0070], + [-0.0145, 0.0260, -0.0085, ..., 0.0077, 0.0102, -0.0166], + [ 0.0094, -0.0142, -0.0037, ..., 0.0051, 0.0069, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0070, 0.0130, -0.0076, ..., -0.0278, 0.0240, -0.0126], + [ 0.0035, 0.0187, -0.0160, ..., -0.0317, -0.0032, 0.0173], + [ 0.0062, 0.0310, -0.0092, ..., -0.0226, 0.0030, 0.0334], + ..., + [-0.0079, 0.0186, 0.0038, ..., 0.0228, 0.0069, 0.0065], + [ 0.0271, 0.0117, 0.0354, ..., -0.0076, 0.0198, 0.0145], + [ 0.0282, 0.0171, 0.0317, ..., -0.0111, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0114, -0.0265, -0.0110, ..., 0.0049, 0.0037, 0.0136], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0085], + [-0.0067, -0.0190, -0.0045, ..., 0.0166, -0.0135, 0.0138], + ..., + [-0.0200, -0.0012, 0.0061, ..., -0.0339, 0.0361, -0.0025], + [ 0.0037, -0.0250, 0.0153, ..., 0.0006, -0.0162, 0.0220], + [-0.0078, 0.0028, 0.0205, ..., -0.0064, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0053, 0.0240, -0.0217, ..., 0.0172, 0.0048, 0.0114], + [-0.0386, 0.0103, 0.0075, ..., -0.0042, 0.0322, 0.0086], + [-0.0427, 0.0040, 0.0035, ..., 0.0073, 0.0110, -0.0137], + ..., + [-0.0087, -0.0041, -0.0264, ..., -0.0586, -0.0630, -0.0132], + [ 0.0045, 0.0287, 0.0420, ..., -0.0311, -0.0090, -0.0289], + [-0.0107, -0.0610, 0.0162, ..., -0.0291, -0.0166, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0121, -0.0251, 0.0211, ..., 0.0199, -0.0084, -0.0112], + [ 0.0112, -0.0099, -0.0327, ..., 0.0447, 0.0076, 0.0074], + [ 0.0035, -0.0175, 0.0066, ..., -0.0062, -0.0049, -0.0065], + ..., + [ 0.0295, 0.0045, -0.0104, ..., 0.0093, -0.0160, -0.0288], + [ 0.0083, 0.0253, -0.0177, ..., 0.0094, 0.0074, 0.0080], + [-0.0129, 0.0027, 0.0137, ..., -0.0128, 0.0277, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0131, 0.0047, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0068, 0.0234, 0.0065, ..., 0.0464, 0.0064, 0.0309], + ..., + [-0.0117, -0.0175, 0.0327, ..., 0.0132, 0.0044, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, 0.0059, -0.0444, ..., 0.0060, -0.0283, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0082, -0.0099, 0.0146, ..., -0.0742, -0.0064, 0.0161], + [-0.0076, -0.0131, -0.0075, ..., 0.0064, 0.0026, 0.0238], + [ 0.0189, 0.0221, -0.0088, ..., 0.0050, 0.0408, 0.0052], + ..., + [ 0.0059, 0.0247, 0.0396, ..., 0.0064, -0.0304, -0.0183], + [ 0.0299, -0.0114, -0.0417, ..., -0.0048, -0.0376, -0.0044], + [-0.0210, 0.0168, -0.0026, ..., 0.0049, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0069, 0.0242, 0.0062, ..., -0.0245, 0.0078, 0.0002], + [ 0.0190, -0.0125, 0.0332, ..., -0.0146, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0068, ..., -0.0059, -0.0092, 0.0182], + ..., + [-0.0430, -0.0403, -0.0081, ..., 0.0135, -0.0140, 0.0046], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0094, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0042, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0092, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0067, ..., 0.0101, -0.0223, 0.0062], + ..., + [ 0.0192, 0.0292, 0.0059, ..., 0.0176, -0.0197, 0.0046], + [ 0.0265, 0.0063, 0.0054, ..., 0.0286, -0.0021, -0.0267], + [-0.0184, -0.0088, -0.0048, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0084, -0.0159, 0.0082, ..., 0.0198, 0.0056, 0.0277], + [ 0.0172, 0.0041, -0.0057, ..., 0.0048, -0.0049, 0.0057], + [ 0.0036, 0.0184, 0.0113, ..., 0.0099, -0.0056, -0.0074], + ..., + [ 0.0048, -0.0111, 0.0035, ..., -0.0096, -0.0053, 0.0122], + [-0.0427, -0.0391, -0.0192, ..., 0.0091, -0.0238, -0.0078], + [ 0.0055, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0084, -0.0161, 0.0123], + [-0.0081, 0.0060, 0.0312, ..., 0.0053, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0071, -0.0014, 0.0278], + ..., + [ 0.0170, -0.0085, 0.0471, ..., 0.0067, 0.0078, -0.0073], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0076, -0.0203], + [ 0.0053, 0.0048, -0.0081, ..., -0.0095, -0.0086, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[-0.0058, -0.0156, 0.0284, ..., -0.0104, 0.0135, -0.0044], + [ 0.0152, 0.0332, 0.0337, ..., 0.0223, -0.0165, -0.0044], + [ 0.0055, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [-0.0056, -0.0019, 0.0280, ..., -0.0076, -0.0078, 0.0140], + [-0.0083, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0091, ..., 0.0067, 0.0167, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0042, ..., -0.0238, -0.0103, 0.0119], + [-0.0250, -0.0162, 0.0034, ..., -0.0081, 0.0002, -0.0369], + [ 0.0051, 0.0141, 0.0079, ..., -0.0452, -0.0038, 0.0120], + ..., + [ 0.0148, -0.0141, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0101, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0142, 0.0041, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[ 0.0028, -0.0327, -0.0190, ..., 0.0064, -0.0052, -0.0209], + [ 0.0713, 0.0152, -0.0052, ..., -0.0049, 0.0186, 0.0027], + [-0.0227, -0.0092, 0.0188, ..., -0.0122, 0.0193, 0.0053], + ..., + [ 0.0116, 0.0464, 0.0085, ..., 0.0248, 0.0168, 0.0047], + [ 0.0381, -0.0038, 0.0071, ..., -0.0031, -0.0107, -0.0054], + [ 0.0292, 0.0057, 0.0209, ..., 0.0056, 0.0234, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0056, 0.0212, ..., 0.0305, 0.0137, 0.0105], + [-0.0078, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0045], + ..., + [ 0.0084, -0.0275, 0.0225, ..., 0.0043, -0.0189, -0.0107], + [-0.0128, 0.0053, -0.0058, ..., -0.0079, 0.0064, 0.0211], + [ 0.0015, -0.0073, -0.0063, ..., -0.0109, 0.0264, 0.0059]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[-0.0049, -0.0075, -0.0108, ..., 0.0186, -0.0216, -0.0047], + [ 0.0048, 0.0123, -0.0231, ..., 0.0123, -0.0112, -0.0183], + [ 0.0107, 0.0103, -0.0057, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0125, 0.0154, ..., 0.0540, 0.0415, 0.0097], + [ 0.0025, 0.0087, -0.0137, ..., -0.0030, -0.0189, -0.0237], + [ 0.0059, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0076, 0.0359, ..., -0.0084, -0.0120, -0.0049], + [ 0.0190, -0.0059, 0.0298, ..., 0.0126, -0.0050, -0.0200], + [ 0.0112, -0.0110, 0.0177, ..., -0.0248, 0.0229, 0.0034], + ..., + [-0.0081, 0.0044, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0044, -0.0291, ..., 0.0206, 0.0304, -0.0085], + [-0.0080, 0.0245, -0.0068, ..., -0.0413, 0.0216, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0078, -0.0157, 0.0043, ..., -0.0052, 0.0228, -0.0339], + [-0.0713, -0.0153, -0.0168, ..., -0.0217, -0.0183, -0.0073], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, -0.0063], + ..., + [-0.0027, -0.0069, 0.0258, ..., -0.0310, 0.0091, -0.0361], + [ 0.0063, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [-0.0071, 0.0157, -0.0069, ..., 0.0075, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0077, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0106, ..., 0.0104, -0.0210, 0.0148], + [-0.0286, -0.0267, -0.0037, ..., 0.0129, -0.0059, 0.0087], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [-0.0062, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0129, 0.0206, ..., 0.0069, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0152, -0.0092, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0045, 0.0076, ..., -0.0046, 0.0110, 0.0598], + [-0.0062, 0.0267, -0.0053, ..., 0.0284, -0.0045, -0.0364], + ..., + [ 0.0045, 0.0078, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0017, -0.0087, ..., 0.0211, 0.0137, -0.0075], + [ 0.0101, 0.0236, -0.0206, ..., -0.0136, -0.0063, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0067, 0.0294, ..., 0.0040, 0.0270, 0.0065], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0066, -0.0168], + [ 0.0201, -0.0121, 0.0126, ..., -0.0381, -0.0247, -0.0059], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0078], + [-0.0194, 0.0094, -0.0092, ..., -0.0231, -0.0069, -0.0466], + [ 0.0327, 0.0284, -0.0044, ..., 0.0145, 0.0142, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, -0.0021, -0.0167, ..., -0.0037, 0.0170, 0.0107], + [ 0.0019, -0.0031, 0.0295, ..., -0.0172, 0.0040, 0.0135], + [ 0.0193, 0.0236, -0.0125, ..., 0.0038, -0.0369, 0.0239], + ..., + [ 0.0151, -0.0413, -0.0124, ..., -0.0354, 0.0065, -0.0320], + [-0.0266, -0.0063, -0.0327, ..., -0.0050, 0.0058, 0.0073], + [ 0.0298, 0.0356, -0.0192, ..., 0.0042, 0.0008, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0045, 0.0058, 0.0317, ..., 0.0062, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0071, -0.0096], + [-0.0166, 0.0064, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0041, 0.0260, ..., 0.0159, -0.0035, -0.0474], + [-0.0064, 0.0079, 0.0076, ..., 0.0140, 0.0557, 0.0052], + [-0.0203, 0.0273, -0.0040, ..., -0.0330, -0.0038, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-3.3203e-02, -1.2390e-02, -3.8757e-03, ..., 6.3782e-03, + -9.3384e-03, -2.7710e-02], + [ 4.9744e-03, -3.4637e-03, -5.5237e-03, ..., 6.5002e-03, + 1.4038e-02, -1.6113e-02], + [-4.6997e-03, 1.6846e-02, -7.6599e-03, ..., -9.1553e-03, + 1.2390e-02, 2.0447e-03], + ..., + [ 3.9795e-02, 4.5166e-02, 2.5146e-02, ..., -5.8289e-03, + -3.2959e-02, 4.1199e-03], + [-6.2012e-02, 3.5248e-03, 3.7109e-02, ..., 4.1016e-02, + 9.2506e-05, -1.7334e-02], + [ 1.2634e-02, 5.7678e-03, -2.6978e-02, ..., 4.9438e-03, + 3.8818e-02, -1.0376e-02]], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0013, 0.0245], + [-0.0449, 0.0019, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, 0.0062, ..., 0.0041, 0.0288, 0.0124], + [ 0.0330, 0.0038, 0.0060, ..., 0.0299, 0.0142, -0.0483], + [-0.0166, 0.0077, 0.0051, ..., -0.0062, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0045, -0.0025, 0.0527], + [-0.0002, 0.0403, 0.0112, ..., 0.0417, 0.0105, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0096, -0.0096, 0.0327], + [ 0.0142, 0.0111, -0.0069, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0124, 0.0281, ..., -0.0242, -0.0127, -0.0047]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0042], + [ 0.0063, -0.0045, -0.0286, ..., 0.0056, 0.0172, 0.0106], + [ 0.0393, 0.0400, 0.0039, ..., 0.0025, 0.0082, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0107, 0.0097, 0.0150], + [-0.0057, 0.0039, 0.0077, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0078, ..., 0.0092, -0.0114, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0071, 0.0041, ..., -0.0220, -0.0145, -0.0013], + [-0.0166, 0.0189, -0.0047, ..., 0.0096, 0.0020, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0054, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0057, ..., 0.0047, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0082, 0.0071, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0122, 0.0289, 0.0125, ..., 0.0135, 0.0070, -0.0104], + [ 0.0188, -0.0197, 0.0010, ..., 0.0284, -0.0342, -0.0057], + [-0.0525, -0.0215, 0.0154, ..., -0.0270, 0.0081, 0.0254], + ..., + [-0.0069, 0.0087, 0.0281, ..., 0.0129, -0.0064, -0.0048], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0045, 0.0038], + [-0.0273, -0.0747, 0.0208, ..., -0.0044, -0.0044, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0048, 0.0150, 0.0334, ..., 0.0205, -0.0273, -0.0037], + [-0.0165, 0.0111, -0.0135, ..., 0.0106, -0.0071, 0.0219], + [ 0.0063, 0.0227, -0.0054, ..., 0.0128, -0.0077, 0.0150], + ..., + [-0.0347, -0.0148, 0.0053, ..., -0.0046, -0.0042, -0.0094], + [-0.0077, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0069, -0.0500, -0.0232, ..., 0.0320, 0.0057, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0056, -0.0364], + [ 0.0220, -0.0026, -0.0112, ..., -0.0052, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, 0.0039, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0064, 0.0192, -0.0393], + [-0.0102, -0.0244, 0.0275, ..., 0.0029, 0.0042, 0.0081], + [-0.0170, -0.0427, 0.0140, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0040, -0.0378, -0.0359, ..., 0.0552, -0.0062, -0.0160], + [ 0.0039, 0.0143, 0.0134, ..., 0.0049, -0.0153, 0.0354], + [ 0.0095, -0.0038, 0.0065, ..., -0.0046, -0.0249, -0.0136], + ..., + [ 0.0081, -0.0124, 0.0050, ..., -0.0095, -0.0085, 0.0047], + [-0.0051, 0.0439, 0.0166, ..., 0.0248, 0.0131, -0.0187], + [ 0.0175, 0.0059, 0.0041, ..., 0.0410, -0.0096, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0104, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0141, -0.0050, ..., 0.0052, -0.0244, 0.0356], + [ 0.0063, 0.0165, -0.0200, ..., -0.0019, 0.0237, -0.0078], + ..., + [-0.0197, -0.0049, 0.0386, ..., 0.0181, 0.0112, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, 0.0041, -0.0374], + [-0.0086, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0066, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0135, 0.0114, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, 0.0045, -0.0041, ..., 0.0050, -0.0055, 0.0337], + ..., + [ 0.0422, 0.0033, 0.0352, ..., -0.0304, -0.0049, -0.0204], + [ 0.0302, 0.0156, 0.0099, ..., -0.0078, -0.0254, -0.0245], + [-0.0040, -0.0060, -0.0039, ..., 0.0037, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0052, 0.0266, ..., 0.0154, 0.0187, 0.0047], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0051, ..., -0.0172, -0.0091, 0.0231], + ..., + [-0.0295, -0.0137, 0.0038, ..., 0.0043, -0.0044, 0.0099], + [-0.0049, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0078], + [-0.0040, -0.0048, -0.0259, ..., -0.0039, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0079, 0.0078, -0.0140, ..., 0.0242, -0.0070, -0.0386], + [-0.0140, 0.0079, -0.0097, ..., -0.0369, -0.0156, 0.0157], + [-0.0267, 0.0078, -0.0057, ..., -0.0126, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0007, 0.0144, ..., 0.0115, 0.0042, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., -0.0046, 0.0042, -0.0137], + [-0.0219, -0.0396, 0.0041, ..., -0.0058, -0.0018, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0034, 0.0168, 0.0194, ..., 0.0349, -0.0420, 0.0017], + [ 0.0042, 0.0273, 0.0145, ..., -0.0184, -0.0039, 0.0154], + [ 0.0145, -0.0036, -0.0072, ..., 0.0245, 0.0359, 0.0059], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0046, 0.0033, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0070, -0.0312, -0.0457], + [ 0.0050, 0.0054, -0.0053, ..., 0.0049, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0015, 0.0123, -0.0029, ..., 0.0124, -0.0038, 0.0066], + [-0.0277, -0.0036, -0.0181, ..., -0.0027, -0.0034, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0042, 0.0513, -0.0110], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0111], + [-0.0282, -0.0053, 0.0121, ..., -0.0051, -0.0039, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0039, -0.0128, ..., 0.0140, 0.0308, 0.0058], + [ 0.0119, -0.0332, -0.0320, ..., -0.0078, -0.0140, -0.0029], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, -0.0028, -0.0311, ..., 0.0120, 0.0061, 0.0081], + [ 0.0048, 0.0198, 0.0142, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, -0.0046, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0091, -0.0036, 0.0056, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0018, ..., 0.0302, -0.0024, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0039, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0025, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0049, -0.0222, 0.0262], + ..., + [-0.0147, 0.0058, 0.0260, ..., 0.0051, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0057, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0076, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0113], + [-0.0278, 0.0044, -0.0374, ..., 0.0276, -0.0038, 0.0151], + [ 0.0089, 0.0237, -0.0067, ..., 0.0527, -0.0070, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0068, -0.0095], + [-0.0075, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0044, -0.0128, ..., 0.0049, -0.0371, 0.0017], + [ 0.0075, -0.0126, 0.0049, ..., -0.0204, 0.0041, 0.0056], + [-0.0110, -0.0195, 0.0056, ..., -0.0542, -0.0092, -0.0053], + ..., + [-0.0205, -0.0085, -0.0041, ..., -0.0199, 0.0041, 0.0061], + [-0.0220, -0.0302, -0.0099, ..., 0.0104, 0.0369, 0.0043], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[-0.0023, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0047, 0.0049, 0.0238, ..., -0.0145, -0.0320, 0.0037], + [-0.0193, 0.0364, -0.0048, ..., -0.0064, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0048, -0.0232, -0.0464], + [-0.0109, -0.0298, -0.0195, ..., 0.0114, 0.0312, 0.0156], + [-0.0074, -0.0117, -0.0195, ..., 0.0043, 0.0106, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0048, -0.0239, -0.0123, ..., -0.0045, 0.0243, -0.0396], + [-0.0101, 0.0155, -0.0056, ..., -0.0277, -0.0038, 0.0037], + ..., + [ 0.0117, -0.0057, 0.0515, ..., 0.0327, -0.0120, -0.0066], + [ 0.0041, 0.0170, 0.0074, ..., -0.0048, 0.0280, -0.0210], + [ 0.0036, 0.0068, 0.0057, ..., 0.0153, -0.0077, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0061, 0.0081, -0.0077, ..., -0.0588, 0.0187, -0.0064], + [-0.0102, 0.0199, -0.0077, ..., 0.0080, -0.0332, 0.0040], + ..., + [ 0.0078, -0.0464, 0.0277, ..., -0.0145, 0.0198, -0.0078], + [ 0.0162, 0.0038, 0.0090, ..., -0.0109, 0.0081, 0.0297], + [-0.0269, 0.0135, 0.0164, ..., 0.0079, 0.0176, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0042, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, -0.0026, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, 0.0048, -0.0201, ..., 0.0330, 0.0309, -0.0051], + ..., + [ 0.0039, 0.0123, 0.0153, ..., 0.0039, -0.0039, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0040, 0.0175], + [-0.0244, -0.0059, -0.0300, ..., 0.0049, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0045, -0.0041, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0038, ..., 0.0310, 0.0088, -0.0039], + [ 0.0025, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0110, -0.0237, ..., -0.0104, 0.0157, 0.0078], + [-0.0070, 0.0205, 0.0143, ..., 0.0262, -0.0195, 0.0095], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0078]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0024, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0022, 0.0034, -0.0093], + [ 0.0017, -0.0010, 0.0398, ..., -0.0304, -0.0028, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0039], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0051, 0.0291], + [ 0.0132, 0.0038, -0.0339, ..., 0.0077, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, -0.0027], + [-0.0166, -0.0036, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0061, -0.0032, ..., 0.0240, 0.0036, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0019], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0007, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0067, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0040], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, 0.0040, -0.0039], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, -0.0039, ..., -0.0164, -0.0039, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0026, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0025, 0.0226], + ..., + [ 0.0053, -0.0231, -0.0255, ..., -0.0052, 0.0295, -0.0058], + [-0.0255, 0.0513, 0.0039, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0040, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0056, 0.0078, -0.0364, ..., -0.0175, 0.0306, -0.0075], + [-0.0051, -0.0195, -0.0016, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0070, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0078, -0.0193, -0.0140], + [-0.0166, 0.0059, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0138, 0.0082, 0.0146, ..., -0.0081, 0.0211, -0.0270], + [ 0.0063, -0.0077, -0.0234, ..., 0.0089, -0.0055, -0.0165], + [ 0.0225, -0.0251, 0.0047, ..., 0.0072, 0.0236, -0.0036], + ..., + [ 0.0186, 0.0148, -0.0053, ..., 0.0272, -0.0062, -0.0086], + [ 0.0064, 0.0029, 0.0078, ..., -0.0260, 0.0267, 0.0147], + [ 0.0101, -0.0044, 0.0212, ..., 0.0195, -0.0157, 0.0114]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0084, -0.0270, 0.0251, ..., 0.0098, 0.0055, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.2.proj.1.weight': tensor([[ 0.0036, 0.0231, 0.0918, ..., -0.0571, 0.0250, -0.0284], + [-0.0237, 0.0006, -0.1128, ..., 0.0193, -0.0238, 0.0518], + [-0.0087, 0.1279, -0.0123, ..., -0.0220, 0.0184, -0.0349], + ..., + [ 0.0146, 0.0055, -0.0076, ..., 0.0234, -0.0280, 0.0082], + [-0.0237, -0.0608, 0.0603, ..., 0.0332, 0.0708, 0.0505], + [ 0.0146, -0.0229, -0.0050, ..., 0.0249, 0.0420, -0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_proj.weight': tensor([[-0.0093, 0.0068, 0.0118, ..., -0.0078, -0.0047, 0.0182], + [ 0.0018, 0.0015, -0.0032, ..., -0.0032, -0.0005, -0.0092], + [-0.0064, 0.0067, -0.0002, ..., 0.0002, -0.0104, -0.0227], + ..., + [ 0.0010, -0.0172, -0.0045, ..., 0.0082, 0.0030, -0.0064], + [-0.0014, 0.0045, -0.0045, ..., -0.0121, -0.0164, -0.0048], + [ 0.0080, -0.0050, -0.0016, ..., 0.0063, 0.0061, 0.0010]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_norm.weight': tensor([ 1.4062e-01, -4.2480e-02, 8.6426e-02, 2.3145e-01, -3.4485e-03, + -8.3160e-04, -1.3933e-06, -1.8794e-21, 6.6174e-22, -2.0117e-20, + -9.3408e-33, 3.3352e-21, -6.1512e-05, -1.1942e-23, -2.9565e-25, + 2.0954e-30, 1.8597e-05, 2.6484e-09, -3.7517e-12, -1.8406e-04, + 2.4738e-10, -4.5528e-21, 4.1504e-02, -3.2616e-04, -1.0742e-01, + 5.8594e-01, 5.4932e-02, -7.3853e-03, 4.9805e-02, 9.8633e-02, + -6.3019e-19, -1.4095e-21, 1.0757e-03, -9.5367e-05, 1.2346e-13, + -2.4033e-04, 3.4106e-12, -7.1086e-25, 1.1529e-23, -2.0292e-24, + -2.3588e-25, -6.0988e-26, 1.1228e-25, -6.5308e-03, 4.2969e-02, + -1.4771e-02, 2.9785e-02, 1.2354e-01, 3.4180e-01, 7.8125e-03, + -4.6082e-03, 5.3406e-03, 4.4556e-03, 1.2005e-10, 2.8922e-10, + 1.4095e-17, -4.8307e-22, -1.0503e-18, 3.0494e-26, -2.6496e-24, + 5.1670e-29, -8.8938e-20, 2.1487e-25, 2.4536e-14, 2.7148e-01, + 2.2559e-01, 3.7109e-01, -4.9316e-02, 2.6855e-03, 7.1335e-04, + 1.0282e-06, 4.3675e-22, 1.2811e-20, 5.7175e-20, -1.3482e-31, + -7.3123e-22, 3.6163e-03, -7.3412e-24, -2.1306e-26, 6.2320e-29, + 1.9226e-03, -9.7316e-11, 4.0927e-11, -3.8452e-03, 3.3253e-12, + -6.6746e-19, -3.9978e-03, -1.2360e-03, 5.7812e-01, 1.1719e-01, + 5.0391e-01, 9.2773e-02, -5.8594e-03, 9.0408e-04, 1.6263e-18, + -5.2675e-21, 3.0060e-03, 4.6730e-05, 4.2188e-14, 3.9978e-03, + -5.7116e-10, -1.4217e-24, -8.4011e-25, -2.8290e-22, -4.2914e-28, + -2.5022e-23, 8.5591e-29, 1.4404e-02, 5.4626e-03, -6.0425e-03, + 4.1211e-01, 2.7734e-01, 2.0410e-01, -1.1047e-02, -5.2185e-03, + 7.8125e-03, 4.6692e-03, -4.8885e-11, 1.4461e-10, -9.1480e-19, + 4.1558e-21, 1.2197e-18, 3.3120e-26, -3.9911e-23, -1.2572e-29, + -3.2591e-22, -1.7707e-24, 8.5038e-11], dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.self_attn.k_proj.weight': tensor([[-0.0053, 0.0036, -0.0114, ..., -0.0015, -0.0072, 0.0092], + [-0.0233, -0.0015, 0.0042, ..., -0.0058, -0.0069, -0.0112], + [-0.0036, -0.0111, -0.0016, ..., 0.0052, 0.0021, 0.0089], + ..., + [ 0.0049, -0.0039, -0.0018, ..., -0.0025, 0.0008, 0.0016], + [ 0.0049, 0.0009, 0.0007, ..., -0.0044, 0.0020, 0.0024], + [ 0.0040, 0.0023, 0.0050, ..., -0.0034, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.k_norm.weight': tensor([ 2.4121e-01, 1.2598e-01, 6.2012e-02, 2.0898e-01, 3.3417e-03, + -1.1292e-03, 1.4082e-06, -1.9852e-21, 2.2896e-21, 2.8164e-20, + 1.9548e-32, 3.2955e-21, -9.9659e-05, -1.1581e-23, -2.9404e-25, + 2.7364e-30, -7.3016e-07, 2.6921e-09, 2.0464e-12, 1.9360e-04, + -2.2555e-10, 2.7926e-21, 3.7354e-02, -9.9182e-04, 7.6172e-02, + 5.8594e-01, 2.2852e-01, 8.9844e-02, 6.2256e-03, 6.8970e-03, + 2.5072e-19, 9.3306e-22, -3.7003e-04, -9.8228e-05, 1.4122e-13, + -1.0824e-04, -3.0695e-11, -1.3248e-25, -1.1684e-23, 1.0792e-24, + -2.2941e-25, -1.1551e-25, -1.1228e-25, 1.0681e-02, 4.4678e-02, + -1.1536e-02, 2.2852e-01, 2.0996e-01, 3.7109e-01, 1.2085e-02, + -4.2725e-03, -7.8125e-03, 7.6904e-03, -1.2824e-10, 3.1650e-10, + -7.5352e-18, -1.7999e-21, 1.3281e-18, 4.2005e-26, -2.7271e-24, + -5.4037e-29, -8.6821e-20, -2.1487e-25, 2.6867e-14, -1.0547e-01, + 2.2168e-01, 3.7305e-01, -1.5430e-01, -2.1667e-03, -9.6321e-05, + -1.0431e-06, -6.3693e-23, 1.2705e-20, 5.5904e-20, 1.3173e-31, + 1.0059e-21, -3.5858e-03, 6.2814e-24, -2.0700e-26, 6.6659e-29, + 1.9302e-03, 2.0191e-10, 4.2292e-11, -3.8452e-03, -2.7427e-12, + -6.5391e-19, -7.8735e-03, -2.8229e-04, 5.7812e-01, -1.3477e-01, + 4.9609e-01, 7.0801e-03, 4.9072e-02, 9.4727e-02, -1.8431e-18, + -5.2675e-21, -3.1891e-03, 2.1338e-05, -1.9207e-14, -3.9978e-03, + 5.8208e-10, -1.4282e-24, 3.2829e-24, -2.7132e-22, 1.6787e-27, + -1.5303e-23, 2.9819e-28, 9.4604e-03, 2.0294e-03, 1.0559e-02, + 4.0430e-01, 2.3340e-01, 1.8555e-01, 8.2397e-03, 5.9814e-03, + 4.6387e-03, -1.2054e-03, 1.6257e-11, 2.6489e-11, -8.4026e-18, + 3.0308e-21, -2.2192e-19, -2.1104e-26, 3.8877e-23, -1.0748e-29, + 6.0880e-21, -1.8482e-24, -8.4583e-11], dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.self_attn.v_proj.weight': tensor([[-0.0007, 0.0037, -0.0014, ..., 0.0045, 0.0013, -0.0028], + [ 0.0083, 0.0014, -0.0035, ..., -0.0039, 0.0026, 0.0009], + [ 0.0136, 0.0058, 0.0026, ..., -0.0038, -0.0015, 0.0028], + ..., + [-0.0036, -0.0049, 0.0043, ..., 0.0056, -0.0034, -0.0077], + [-0.0043, -0.0018, -0.0001, ..., 0.0055, -0.0032, -0.0046], + [-0.0015, 0.0055, -0.0029, ..., 0.0037, 0.0025, -0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.output_proj.weight': tensor([[ 8.4229e-03, -3.6133e-02, -1.4099e-02, ..., -6.5918e-03, + 5.1270e-03, 5.0659e-03], + [ 6.6223e-03, -7.2327e-03, 5.8594e-03, ..., -2.6703e-03, + 1.8311e-03, 3.2043e-03], + [-3.2196e-03, 5.2214e-05, 7.5684e-03, ..., 4.5776e-03, + 4.5166e-03, 2.8839e-03], + ..., + [ 2.7008e-03, 5.2490e-03, 7.0801e-03, ..., -4.4250e-03, + -6.4697e-03, -6.5002e-03], + [-2.2125e-03, 5.1270e-03, -6.3477e-03, ..., -9.4604e-03, + 8.1787e-03, 4.3640e-03], + [-4.2114e-03, 1.0132e-02, 3.9368e-03, ..., 4.6997e-03, + 4.4861e-03, 4.4556e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_proj.weight': tensor([[ 0.0085, 0.0082, -0.0034, ..., -0.0069, 0.0062, -0.0014], + [ 0.0073, -0.0261, -0.0047, ..., 0.0066, -0.0082, 0.0026], + [-0.0121, 0.0070, -0.0002, ..., 0.0057, 0.0209, -0.0048], + ..., + [ 0.0033, 0.0048, -0.0010, ..., -0.0012, -0.0046, 0.0108], + [-0.0066, -0.0033, 0.0006, ..., 0.0024, 0.0063, -0.0033], + [-0.0024, -0.0065, 0.0020, ..., -0.0033, -0.0007, 0.0029]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_norm.weight': tensor([ 9.7275e-04, 1.0315e-02, 3.7107e-10, 2.0752e-03, 2.0294e-03, + 2.3723e-05, -1.9550e-05, 5.0964e-03, 3.4180e-02, 2.6001e-02, + 6.8918e-07, 2.4536e-02, 5.5469e-01, 7.4768e-03, 8.4229e-03, + 5.3711e-03, 4.6875e-01, 1.6689e-05, 3.6163e-03, 1.1265e-05, + 5.1498e-04, 4.2725e-03, 5.9814e-02, 1.5991e-02, 1.1963e-02, + 6.3171e-03, 2.5146e-02, -7.5817e-05, 9.9487e-03, 6.7444e-03, + 4.1602e-01, 2.5146e-02, 4.8256e-04, 4.2188e-01, 5.3125e-01, + 1.9684e-03, 3.4790e-03, 3.4668e-02, 1.1475e-01, 1.4893e-02, + 3.8818e-02, 4.0894e-03, 6.4028e-09, 1.4496e-03, 1.8921e-02, + 2.6855e-02, 1.1719e-02, 9.7656e-03, 1.4954e-03, 9.8348e-06, + 8.5449e-03, -9.7156e-06, -7.1526e-06, 1.6809e-05, -4.6539e-04, + -2.6941e-05, 1.8677e-02, 1.0620e-02, 1.0620e-02, 2.9907e-03, + 4.0054e-04, 1.8716e-05, 8.7280e-03, -1.3530e-05, 1.3828e-04, + 1.4782e-04, 1.7047e-05, 2.5391e-01, 7.4387e-04, 1.7090e-02, + 4.0430e-01, -1.9455e-03, 3.9307e-02, 4.9744e-03, 5.4240e-06, + 1.3351e-04, 9.6436e-03, 4.9591e-04, 2.9449e-03, 1.8188e-02, + 2.0386e-02, 1.6968e-02, 1.9043e-01, 6.1951e-03, 3.4094e-05, + 1.5564e-02, 5.4321e-03, 2.0862e-05, 6.1512e-05, 4.7656e-01, + 7.6599e-03, -8.3447e-05, 1.0193e-02, 3.0100e-06, -5.2691e-05, + 4.3555e-01, 2.8419e-04, 7.9346e-03, -1.8304e-11, 1.0834e-03, + 4.1797e-01, 1.7166e-05, 9.5825e-03, -4.8578e-06, 8.3618e-03, + -3.4571e-05, 2.0409e-04, -2.4033e-04, 2.3804e-02, 1.1597e-02, + 6.1951e-03, -4.7982e-06, 3.1494e-02, 1.1301e-04, 1.1841e-02, + 1.4114e-04, 2.5513e-02, 4.1504e-03, 6.7749e-03, 4.1199e-04, + 2.5272e-05, 1.1169e-02, 8.7280e-03, 6.9885e-03, 1.5198e-02, + 2.6512e-04, 2.6855e-03, 5.3167e-05], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_proj.weight': tensor([[-0.0008, 0.0007, 0.0026, ..., -0.0004, -0.0014, 0.0005], + [-0.0063, -0.0021, -0.0051, ..., -0.0028, -0.0053, -0.0034], + [-0.0005, -0.0015, -0.0016, ..., -0.0027, -0.0029, -0.0014], + ..., + [ 0.0026, -0.0025, 0.0018, ..., 0.0011, 0.0024, -0.0006], + [ 0.0010, 0.0005, -0.0018, ..., -0.0025, 0.0009, 0.0019], + [ 0.0035, 0.0019, -0.0023, ..., -0.0041, -0.0006, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_norm.weight': tensor([-9.7275e-04, 1.0254e-02, 3.6016e-10, 1.8921e-03, 2.0142e-03, + 2.7061e-05, 1.9550e-05, 5.0659e-03, 3.3203e-02, 2.6123e-02, + -7.1898e-07, 2.4902e-02, 5.5859e-01, 7.4463e-03, 7.9346e-03, + 5.3406e-03, 4.6875e-01, 1.7524e-05, 3.5248e-03, -2.5392e-05, + 5.7983e-04, 4.3030e-03, 5.9814e-02, 1.5991e-02, 1.1780e-02, + 6.3782e-03, 2.5024e-02, -7.5817e-05, 9.3384e-03, 6.7444e-03, + 4.1602e-01, 2.5024e-02, 4.5395e-04, 4.2188e-01, 5.3125e-01, + 1.8234e-03, 3.5553e-03, 3.4912e-02, 1.1572e-01, 1.4893e-02, + 3.8818e-02, 4.1199e-03, -6.2282e-09, 1.2131e-03, 1.8555e-02, + 2.6733e-02, 1.1719e-02, 9.7656e-03, 1.3580e-03, 9.8944e-06, + 8.4839e-03, -1.0431e-05, -7.1526e-06, 1.9789e-05, 4.6539e-04, + 1.3947e-05, 1.8433e-02, 1.0742e-02, 1.0925e-02, 2.9907e-03, + -2.8419e-04, 1.8716e-05, 8.4229e-03, -1.3888e-05, 8.6308e-05, + 1.8692e-04, 1.6928e-05, 2.5391e-01, 6.4468e-04, 1.7090e-02, + 4.0430e-01, -1.9455e-03, 3.7842e-02, 4.9438e-03, -1.7956e-06, + 1.6117e-04, 8.6670e-03, 4.7112e-04, 2.9297e-03, 1.8188e-02, + 2.0264e-02, 1.7090e-02, 1.9043e-01, 6.0425e-03, 2.6345e-05, + 1.5564e-02, 5.8289e-03, 2.2173e-05, 5.7459e-05, 4.7656e-01, + 7.6599e-03, -9.9182e-05, 9.8877e-03, 2.2948e-06, -5.4359e-05, + 4.3555e-01, -1.0967e-04, 7.9956e-03, 1.8304e-11, 1.0605e-03, + 4.1797e-01, 1.6928e-05, 1.0498e-02, -7.1526e-06, 8.3618e-03, + -3.5048e-05, 2.2697e-04, -2.2984e-04, 2.3438e-02, 1.1536e-02, + 6.2866e-03, -4.2617e-06, 3.1128e-02, 1.1396e-04, 1.2024e-02, + -1.0920e-04, 2.5513e-02, 4.2725e-03, 7.2021e-03, 4.2534e-04, + 2.5153e-05, 1.1292e-02, 8.7280e-03, 6.9885e-03, 1.4771e-02, + 1.6975e-04, 2.7466e-03, -4.8637e-05], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.v_proj.weight': tensor([[-0.0104, 0.0016, 0.0097, ..., 0.0032, 0.0028, -0.0110], + [-0.0243, -0.0013, 0.0063, ..., -0.0010, -0.0076, 0.0008], + [-0.0132, 0.0032, 0.0044, ..., 0.0031, 0.0153, -0.0128], + ..., + [ 0.0117, 0.0008, 0.0018, ..., -0.0023, 0.0064, 0.0181], + [-0.0046, -0.0002, 0.0117, ..., -0.0018, -0.0161, -0.0034], + [-0.0216, -0.0033, 0.0044, ..., -0.0035, -0.0064, -0.0221]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.output_proj.weight': tensor([[ 0.0051, -0.0060, -0.0038, ..., -0.0053, -0.0006, 0.0098], + [ 0.0043, -0.0039, -0.0088, ..., 0.0214, -0.0043, -0.0114], + [-0.0044, 0.0049, 0.0037, ..., 0.0083, 0.0037, 0.0073], + ..., + [-0.0089, -0.0048, 0.0079, ..., -0.0076, -0.0035, -0.0005], + [ 0.0046, -0.0064, 0.0063, ..., 0.0055, -0.0054, -0.0113], + [ 0.0150, -0.0101, -0.0053, ..., 0.0051, -0.0061, 0.0074]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_img.weight': tensor([[ 0.0053, 0.0117, 0.0039, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0072, 0.0249, 0.0291], + [-0.0095, -0.0042, -0.0115, ..., -0.0342, -0.0327, 0.0066], + ..., + [ 0.0078, 0.0045, 0.0045, ..., 0.0086, 0.0286, -0.0165], + [ 0.0068, -0.0053, -0.0101, ..., -0.0101, 0.0142, -0.0309], + [ 0.0242, -0.0102, 0.0255, ..., -0.0227, -0.0044, -0.0153]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, 0.0067], + [ 0.0042, 0.0121, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0084, ..., -0.0059, -0.0057, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0037, 0.0049], + [-0.0160, 0.0051, 0.0039, ..., 0.0133, -0.0173, -0.0046], + [-0.0051, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0098]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0048, -0.0364, ..., -0.0225, -0.0142, -0.0056], + [-0.0157, 0.0051, -0.0364, ..., 0.0101, 0.0120, 0.0036], + [ 0.0043, 0.0615, 0.0010, ..., -0.0205, -0.0306, -0.0115], + ..., + [-0.0038, -0.0038, -0.0103, ..., -0.0211, -0.0019, 0.0288], + [ 0.0266, -0.0123, -0.0044, ..., -0.0420, -0.0282, 0.0053], + [ 0.0286, 0.0325, 0.0093, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.mlp.layer1.weight': tensor([[ 0.0073, 0.0172, -0.0126, ..., 0.0037, -0.0081, 0.0063], + [-0.0317, 0.0056, -0.0040, ..., -0.0054, 0.0044, -0.0088], + [-0.0048, -0.0020, 0.0048, ..., 0.0035, 0.0062, -0.0051], + ..., + [-0.0073, -0.0065, -0.0038, ..., -0.0095, 0.0120, 0.0034], + [ 0.0161, 0.0081, 0.0046, ..., -0.0063, 0.0014, 0.0034], + [-0.0178, -0.0041, -0.0038, ..., 0.0040, -0.0090, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.mlp.layer2.weight': tensor([[ 0.0001, -0.0159, 0.0052, ..., -0.0053, 0.0208, -0.0045], + [-0.0054, -0.0058, -0.0045, ..., -0.0073, 0.0044, -0.0053], + [ 0.0075, 0.0032, 0.0086, ..., 0.0044, 0.0024, -0.0157], + ..., + [ 0.0098, -0.0069, -0.0067, ..., 0.0011, 0.0050, -0.0056], + [-0.0063, 0.0007, 0.0046, ..., 0.0077, 0.0239, -0.0121], + [ 0.0059, -0.0151, 0.0050, ..., 0.0119, -0.0114, -0.0085]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0576, -0.0562, -0.0547, ..., 0.0029, 0.0030, 0.0030], + [ 0.0559, 0.0549, 0.0532, ..., -0.0023, -0.0024, -0.0025], + [ 0.0554, 0.0544, 0.0532, ..., -0.0025, -0.0025, -0.0026], + ..., + [ 0.0566, 0.0557, 0.0535, ..., -0.0022, -0.0024, -0.0024], + [ 0.0535, 0.0525, 0.0505, ..., -0.0021, -0.0021, -0.0021], + [ 0.0571, 0.0559, 0.0547, ..., -0.0029, -0.0030, -0.0029]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_self_attn.2.weight': tensor([[ 0.0019, -0.0016, -0.0020, ..., -0.0028, -0.0024, -0.0018], + [ 0.0017, -0.0021, -0.0018, ..., 0.0024, -0.0017, -0.0018], + [ 0.0002, 0.0033, 0.0002, ..., 0.0035, 0.0034, -0.0003], + ..., + [-0.0374, 0.0452, 0.0442, ..., 0.0420, 0.0374, 0.0374], + [-0.0100, 0.0103, 0.0104, ..., 0.0090, 0.0061, 0.0103], + [ 0.0078, -0.0095, -0.0092, ..., -0.0069, -0.0066, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_cross_attn.1.weight': tensor([[-9.0942e-03, -8.6670e-03, 7.8125e-03, ..., -6.0654e-04, + -5.3406e-04, -7.0572e-04], + [-4.1992e-02, -4.0527e-02, -4.0039e-02, ..., 2.9945e-04, + 3.0518e-04, 3.1471e-04], + [ 4.3213e-02, 4.2236e-02, 4.1748e-02, ..., -8.7738e-05, + -6.7711e-05, -9.2506e-05], + ..., + [ 4.1748e-02, 4.0771e-02, 4.0283e-02, ..., -3.7766e-04, + -3.1090e-04, -3.4714e-04], + [ 5.4688e-02, 5.3223e-02, 5.2002e-02, ..., -2.0981e-04, + -1.3924e-04, -1.9932e-04], + [ 3.8574e-02, 3.6865e-02, 3.6377e-02, ..., -4.7386e-06, + 8.0585e-05, 3.0994e-05]], dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0184, -0.0074, 0.0074, ..., 0.0074, 0.0109, 0.0067], + [-0.0044, -0.0042, 0.0042, ..., 0.0042, 0.0062, 0.0042], + [-0.0007, 0.0049, -0.0077, ..., -0.0049, -0.0074, -0.0078], + ..., + [-0.0078, -0.0265, 0.0332, ..., 0.0265, 0.0432, 0.0286], + [ 0.0114, 0.0063, 0.0122, ..., -0.0059, 0.0040, 0.0134], + [-0.0136, 0.0040, -0.0077, ..., -0.0040, -0.0046, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_mlp.1.weight': tensor([[-0.0037, -0.0033, -0.0031, ..., 0.0002, 0.0004, 0.0004], + [ 0.0540, 0.0518, 0.0503, ..., -0.0020, -0.0023, -0.0024], + [ 0.0459, 0.0444, 0.0437, ..., -0.0033, -0.0036, -0.0036], + ..., + [ 0.0884, 0.0854, 0.0845, ..., -0.0005, -0.0007, -0.0008], + [ 0.0035, 0.0035, 0.0036, ..., -0.0002, -0.0003, -0.0002], + [ 0.0654, 0.0635, 0.0618, ..., -0.0010, -0.0010, -0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_mlp.2.weight': tensor([[-0.0004, 0.0125, 0.0136, ..., 0.0120, 0.0003, 0.0098], + [-0.0039, -0.0009, 0.0010, ..., -0.0024, 0.0039, -0.0014], + [-0.0015, 0.0054, 0.0057, ..., 0.0036, 0.0015, 0.0044], + ..., + [ 0.0084, 0.0654, 0.0845, ..., 0.0688, -0.0084, 0.0500], + [ 0.0065, -0.0017, -0.0024, ..., -0.0022, -0.0064, -0.0020], + [-0.0120, -0.0071, -0.0072, ..., -0.0064, 0.0119, -0.0076]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.before_proj.weight': tensor([[ 0.0081, 0.0043, -0.0024, ..., -0.0027, -0.0028, 0.0038], + [-0.0078, -0.0134, -0.0053, ..., 0.0205, 0.0040, 0.0084], + [-0.0042, -0.0051, 0.0022, ..., 0.0023, 0.0043, -0.0065], + ..., + [-0.0031, 0.0089, 0.0085, ..., 0.0024, -0.0022, -0.0054], + [ 0.0143, 0.0198, 0.0146, ..., 0.0020, 0.0107, 0.0045], + [-0.0003, -0.0062, 0.0044, ..., 0.0002, -0.0022, -0.0056]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.before_proj.bias': tensor([ 0.0018, 0.0023, 0.0036, ..., 0.0046, 0.0033, -0.0115], + dtype=torch.bfloat16), 'net.control_blocks_2.0.after_proj.weight': tensor([[-0.0176, 0.0060, -0.0047, ..., -0.0038, -0.0049, -0.0137], + [-0.0106, -0.0037, -0.0068, ..., -0.0107, 0.0072, 0.0188], + [ 0.0019, 0.0009, -0.0058, ..., -0.0017, 0.0006, -0.0014], + ..., + [-0.0062, -0.0015, -0.0006, ..., 0.0024, 0.0011, -0.0017], + [ 0.0145, -0.0029, 0.0059, ..., -0.0047, -0.0109, 0.0077], + [ 0.0077, 0.0045, -0.0050, ..., 0.0039, -0.0055, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.after_proj.bias': tensor([ 0.0300, -0.0136, 0.0002, ..., -0.0012, -0.0262, 0.0337], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_proj.weight': tensor([[-0.0199, 0.0188, 0.0041, ..., 0.0259, -0.0085, 0.0050], + [-0.0106, -0.0162, 0.0114, ..., -0.0073, -0.0008, -0.0393], + [ 0.0139, 0.0078, 0.0051, ..., 0.0129, 0.0022, -0.0157], + ..., + [ 0.0159, 0.0104, -0.0030, ..., 0.0101, -0.0117, -0.0049], + [-0.0114, -0.0149, -0.0050, ..., -0.0016, -0.0167, -0.0042], + [-0.0027, -0.0081, -0.0071, ..., -0.0084, 0.0095, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_norm.weight': tensor([ 3.0664e-01, 9.6875e-01, 5.7129e-02, 7.6562e-01, 9.4922e-01, + -6.4319e-09, -3.0640e-02, 5.4123e-15, -4.0149e-19, 5.0093e-13, + 3.9916e-28, -8.3778e-33, -5.4919e-35, 3.3253e-22, -2.3074e-29, + 3.4907e-34, 1.4941e-01, 2.9325e-05, 1.5068e-04, -1.8921e-02, + 5.1880e-03, -1.3447e-04, 2.6367e-01, 9.4922e-01, 1.1875e+00, + 3.4766e-01, 7.3547e-03, 4.2383e-01, 1.7090e-02, 4.6327e-12, + 1.3281e-17, 4.6629e-15, 5.6839e-04, 1.4343e-03, -3.5095e-04, + 1.3184e-02, 1.4782e-04, 2.6375e-10, -3.8338e-04, 4.2848e-16, + 2.6455e-17, -2.0539e-14, -3.1896e-21, 8.8281e-01, 2.7148e-01, + 2.0117e-01, 4.7852e-01, 1.1406e+00, 1.2402e-01, -3.6926e-03, + -2.1118e-02, 2.0020e-01, 2.8992e-04, 1.8917e-10, -3.8147e-06, + 7.4593e-16, 7.2632e-03, 4.1034e-13, 7.5731e-29, -6.0486e-34, + 3.8556e-35, 1.6515e-15, -1.2494e-20, 1.0532e-33, 3.4766e-01, + 2.5586e-01, 5.8203e-01, 2.0996e-01, 1.1914e-01, -3.4925e-09, + 1.9287e-02, 2.2898e-15, -1.7364e-19, -5.2935e-13, -7.8097e-29, + 8.7269e-34, 2.7836e-35, -9.9262e-23, -1.1107e-27, -1.1059e-34, + 1.4191e-03, 2.7466e-04, 1.6403e-04, -1.2970e-04, 3.0708e-04, + 4.2236e-02, 9.4922e-01, 2.1582e-01, 2.0312e-01, 1.0781e+00, + 1.9824e-01, 1.8945e-01, -4.1504e-03, -4.0075e-12, -3.6592e-18, + 3.8303e-15, 1.0757e-03, 1.6724e-02, -3.8910e-03, -4.3640e-03, + 3.3188e-04, 5.5252e-11, 2.2799e-06, -7.3726e-18, 6.6058e-15, + -5.0848e-14, -5.5318e-24, 2.0312e-01, 1.1484e+00, 1.1953e+00, + 1.3750e+00, 4.0234e-01, 3.8281e-01, 9.8877e-03, -2.1484e-02, + -2.2583e-02, -7.4863e-05, 1.9554e-10, 2.4915e-05, 1.3791e-16, + 1.6846e-02, 1.0747e-13, 2.2404e-28, -6.4398e-34, -1.0391e-35, + -1.3214e-19, 7.1998e-20, 7.7037e-34], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.self_attn.k_proj.weight': tensor([[ 0.0152, 0.0017, -0.0059, ..., -0.0053, 0.0058, -0.0119], + [ 0.0231, 0.0071, 0.0044, ..., -0.0135, -0.0038, -0.0149], + [ 0.0060, 0.0164, -0.0024, ..., -0.0013, 0.0027, 0.0041], + ..., + [-0.0061, 0.0035, -0.0077, ..., 0.0115, 0.0005, -0.0031], + [-0.0100, -0.0090, -0.0032, ..., 0.0115, -0.0021, -0.0073], + [-0.0007, -0.0018, -0.0008, ..., -0.0059, 0.0046, 0.0048]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.k_norm.weight': tensor([ 3.2617e-01, 9.8047e-01, -1.0205e-01, 7.8516e-01, 9.4922e-01, + 2.9686e-09, 2.6245e-02, -5.9119e-15, -3.9980e-19, 5.4357e-13, + 4.5438e-28, -3.2259e-33, 4.4387e-35, 3.3418e-22, -2.1201e-29, + 3.4155e-34, 1.4941e-01, 2.9206e-05, -7.9155e-05, 5.4932e-02, + 5.1880e-03, 8.8215e-05, 2.9492e-01, 9.5312e-01, 1.1719e+00, + 1.4258e-01, 1.4844e-01, 4.6289e-01, 1.4404e-02, -4.5759e-12, + 1.3390e-17, 4.4964e-15, 5.3406e-04, -1.9302e-03, -3.2806e-04, + 4.2236e-02, 1.4687e-04, 2.1009e-10, -2.3651e-04, -4.2848e-16, + 2.4395e-17, -1.9540e-14, -3.1896e-21, 8.7891e-01, 2.5586e-01, + -4.7302e-03, -3.2227e-01, 1.1484e+00, 1.3672e-01, -3.0518e-03, + 2.8809e-02, 1.2402e-01, 1.0633e-04, -1.5280e-10, -3.4720e-06, + -7.4593e-16, 8.4229e-03, 4.1034e-13, 6.9814e-29, 9.3287e-35, + -1.2413e-35, -1.6515e-15, -1.1911e-20, -1.9410e-34, 3.8281e-01, + 3.1250e-01, 5.8594e-01, 1.6895e-01, -9.9609e-02, -4.1036e-09, + -2.2827e-02, -1.5543e-15, 4.5401e-19, -5.7909e-13, 2.1299e-29, + 1.0171e-33, 1.9184e-35, 8.2718e-23, 1.0539e-27, -4.8750e-34, + 3.4332e-04, 3.2663e-05, -1.6594e-04, -1.0920e-04, 3.0899e-04, + 4.2236e-02, 9.4531e-01, 2.0898e-01, 1.9824e-01, 1.0703e+00, + 1.8555e-01, 2.2266e-01, 9.3994e-03, 3.8085e-12, -4.4181e-18, + -3.6915e-15, 1.2054e-03, -1.5747e-02, -3.8910e-03, 3.7598e-02, + 3.2997e-04, 2.8922e-10, 1.6317e-06, -4.7976e-18, 6.6891e-15, + -5.0848e-14, 5.8678e-24, 2.9297e-01, 1.1406e+00, 1.1953e+00, + 1.3750e+00, 1.2891e-01, 3.9648e-01, -9.7656e-03, 4.3030e-03, + -1.6992e-01, 3.4332e-04, -1.7280e-10, -2.5034e-05, -1.3618e-16, + -1.5991e-02, -1.0569e-13, 1.9958e-28, -4.8750e-34, -3.7616e-35, + 3.3475e-18, 7.1574e-20, -4.5139e-34], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.self_attn.v_proj.weight': tensor([[ 8.2397e-03, 5.5847e-03, 7.1106e-03, ..., -2.0630e-02, + 8.9722e-03, 4.8828e-03], + [ 1.1108e-02, 6.4697e-03, 8.4686e-04, ..., -3.6163e-03, + -4.8523e-03, -4.3335e-03], + [-2.0905e-03, -3.7079e-03, 9.7275e-05, ..., -6.8359e-03, + 4.9744e-03, -7.9346e-03], + ..., + [ 3.6926e-03, 8.4229e-03, 1.3428e-02, ..., 2.0508e-02, + 3.6926e-03, -1.7578e-02], + [-1.8555e-02, 5.4321e-03, 1.1169e-02, ..., 3.5889e-02, + -7.6294e-03, 6.4850e-04], + [-5.1880e-03, -4.0588e-03, -5.6152e-03, ..., 1.9531e-02, + -5.6458e-03, 4.4861e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.output_proj.weight': tensor([[-0.0062, -0.0013, -0.0136, ..., 0.0078, -0.0044, -0.0070], + [ 0.0066, 0.0037, 0.0198, ..., 0.0023, -0.0028, 0.0106], + [-0.0044, -0.0047, -0.0019, ..., -0.0060, 0.0150, 0.0024], + ..., + [ 0.0079, 0.0010, -0.0066, ..., -0.0031, -0.0305, -0.0067], + [ 0.0038, -0.0014, -0.0115, ..., -0.0028, -0.0035, -0.0024], + [-0.0033, 0.0103, -0.0242, ..., 0.0129, 0.0017, -0.0036]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_proj.weight': tensor([[ 0.0050, 0.0261, 0.0086, ..., 0.0053, 0.0153, -0.0013], + [ 0.0194, 0.0076, 0.0065, ..., 0.0080, -0.0079, -0.0320], + [ 0.0176, 0.0128, 0.0025, ..., -0.0038, -0.0118, -0.0253], + ..., + [ 0.0061, 0.0032, -0.0038, ..., 0.0105, -0.0020, 0.0168], + [ 0.0058, 0.0090, 0.0104, ..., -0.0047, -0.0013, 0.0248], + [ 0.0175, -0.0170, -0.0084, ..., 0.0095, 0.0043, -0.0243]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_norm.weight': tensor([ 1.8799e-02, 4.7070e-01, 2.3560e-02, 1.0315e-02, 1.2107e-07, + 6.2937e-10, 3.8338e-04, 5.0391e-01, 8.6212e-04, 4.1602e-01, + 1.9165e-02, 1.9287e-02, 3.5095e-03, 1.3000e-02, 2.3315e-02, + 1.3977e-02, 2.6733e-02, -3.8147e-05, -4.0293e-05, 3.5400e-02, + 2.6367e-01, 3.2227e-02, 1.7700e-03, 1.4343e-02, 3.5400e-03, + 4.1008e-05, 1.8716e-05, 4.5410e-02, -6.2883e-06, 3.8281e-01, + 8.7738e-05, -1.3828e-04, 6.3171e-03, -2.5034e-05, 1.0864e-02, + 1.0193e-02, -2.0646e-10, 4.1406e-01, 1.0803e-02, 2.5635e-03, + 7.7148e-02, 2.8534e-03, 9.8877e-03, 1.0490e-04, 7.6904e-03, + 1.3977e-02, 2.9144e-03, 1.7212e-02, 1.4587e-02, 1.4709e-02, + 7.6904e-03, 4.0039e-01, 1.3611e-02, 1.1719e-02, 3.1494e-02, + 4.5538e-05, 9.1309e-02, 1.4709e-02, 1.5564e-02, 1.3672e-02, + 5.4836e-06, 7.6904e-03, 1.1597e-02, -1.4019e-04, 1.1414e-02, + 5.0659e-03, -8.0585e-05, 3.7689e-03, 1.8921e-02, 8.1543e-02, + 5.9128e-04, 6.1646e-03, 5.2979e-02, 4.1748e-02, 2.3956e-03, + 1.2695e-02, 4.0531e-06, 1.8677e-02, 3.1006e-02, 2.0996e-02, + 5.1880e-03, 1.6113e-02, 2.8320e-02, 3.6955e-06, 2.5391e-02, + 9.3994e-03, 3.5400e-03, 3.3203e-02, 6.9275e-03, 2.9602e-03, + 1.6724e-02, 2.0981e-04, 4.2419e-03, 2.1362e-02, 1.6724e-02, + -1.3351e-04, 9.9487e-03, 3.2227e-02, 1.5320e-02, 3.8086e-02, + 3.0762e-02, 2.4780e-02, 1.1169e-02, 6.7711e-05, -4.9360e-08, + 6.4392e-03, 1.8692e-04, 2.0264e-02, 8.1787e-03, 1.2573e-02, + 4.1809e-03, 1.1414e-02, 2.0508e-02, 5.1172e-01, 2.0447e-03, + 9.6436e-03, 1.0376e-02, 5.7068e-03, 7.8735e-03, 7.9590e-02, + 9.2983e-06, 4.5538e-05, 1.6451e-05, 1.2695e-02, 1.5259e-02, + 1.6357e-02, 2.4902e-02, 1.2573e-02], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_proj.weight': tensor([[ 0.0014, -0.0033, -0.0022, ..., -0.0038, 0.0019, -0.0005], + [-0.0085, -0.0035, -0.0021, ..., 0.0098, 0.0134, 0.0070], + [ 0.0055, 0.0036, 0.0045, ..., 0.0072, 0.0049, -0.0034], + ..., + [-0.0057, -0.0049, 0.0025, ..., -0.0029, -0.0009, -0.0032], + [-0.0078, -0.0053, -0.0126, ..., -0.0014, 0.0068, 0.0006], + [-0.0042, -0.0004, -0.0013, ..., 0.0044, -0.0016, -0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_norm.weight': tensor([ 1.9531e-02, 4.7070e-01, 2.3560e-02, 1.0437e-02, 1.2107e-07, + -2.2828e-10, 4.1962e-04, 5.0391e-01, 8.6594e-04, 4.1602e-01, + 2.0020e-02, 1.9287e-02, 3.5858e-03, 1.3000e-02, 2.3193e-02, + 1.3977e-02, 2.7100e-02, -3.8385e-05, -3.9816e-05, 3.5645e-02, + 2.6367e-01, 3.1982e-02, 1.7776e-03, 1.4587e-02, 3.6163e-03, + 4.0054e-05, 1.8597e-05, 4.5410e-02, -3.4124e-06, 3.8281e-01, + 8.7261e-05, -1.3924e-04, 6.1646e-03, -2.4796e-05, 1.0803e-02, + 1.0315e-02, -1.9008e-10, 4.1406e-01, 1.0864e-02, 2.4567e-03, + 7.6172e-02, 2.8839e-03, 9.6436e-03, 1.0633e-04, 7.5989e-03, + 1.3916e-02, 2.8687e-03, 1.7090e-02, 1.4404e-02, 1.5015e-02, + 7.5989e-03, 4.0039e-01, 1.3611e-02, 1.1230e-02, 3.1250e-02, + 4.4346e-05, 9.1309e-02, 1.4771e-02, 1.5259e-02, 1.3794e-02, + 6.3479e-06, 8.0566e-03, 1.1597e-02, -1.4019e-04, 1.1475e-02, + 4.9744e-03, -8.0585e-05, 3.7689e-03, 1.9043e-02, 8.1543e-02, + 5.6839e-04, 6.1340e-03, 5.2979e-02, 4.1748e-02, 2.4261e-03, + 1.2878e-02, 4.9472e-06, 1.8555e-02, 3.0884e-02, 2.1118e-02, + 5.1880e-03, 1.6113e-02, 2.7710e-02, 4.0531e-06, 2.5513e-02, + 9.3994e-03, 3.3875e-03, 3.2959e-02, 7.2632e-03, 3.0365e-03, + 1.6968e-02, 2.1172e-04, 4.3335e-03, 2.1240e-02, 1.6602e-02, + -1.3351e-04, 9.2773e-03, 3.1982e-02, 1.5259e-02, 3.7598e-02, + 3.1128e-02, 2.4780e-02, 1.1230e-02, 6.6280e-05, -1.7229e-08, + 6.4392e-03, 1.9360e-04, 1.9775e-02, 8.3008e-03, 1.2451e-02, + 4.2114e-03, 1.1169e-02, 2.0020e-02, 5.1172e-01, 2.0752e-03, + 9.5215e-03, 1.0315e-02, 5.5847e-03, 7.9956e-03, 7.9590e-02, + 1.3411e-05, 4.4584e-05, 1.6093e-05, 1.3000e-02, 1.5320e-02, + 1.6602e-02, 2.5024e-02, 1.2329e-02], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.v_proj.weight': tensor([[-0.0210, 0.0036, 0.0016, ..., 0.0012, -0.0206, -0.0220], + [-0.0035, 0.0081, 0.0038, ..., 0.0037, 0.0134, 0.0095], + [-0.0422, -0.0024, -0.0076, ..., -0.0079, -0.0286, -0.0074], + ..., + [-0.0376, 0.0016, -0.0150, ..., 0.0014, 0.0220, 0.0054], + [ 0.0082, 0.0015, -0.0062, ..., 0.0018, 0.0128, 0.0084], + [ 0.0112, 0.0050, -0.0072, ..., 0.0006, 0.0069, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.output_proj.weight': tensor([[ 7.2327e-03, 9.3384e-03, 4.6692e-03, ..., -6.8359e-03, + 2.5330e-03, 6.9885e-03], + [-3.9368e-03, -1.3184e-02, 6.2866e-03, ..., -8.1787e-03, + 6.6833e-03, 8.4229e-03], + [ 2.6245e-03, -2.0599e-03, -1.9684e-03, ..., -1.6785e-04, + -6.1417e-04, -7.0190e-03], + ..., + [-6.5918e-03, 6.3896e-05, 4.1809e-03, ..., 5.4626e-03, + 2.7466e-03, 7.8735e-03], + [ 7.6294e-03, 1.1658e-02, -5.0735e-04, ..., -2.6123e-02, + 7.1106e-03, 2.1606e-02], + [ 4.5166e-03, 4.0588e-03, -1.7456e-02, ..., -1.0742e-02, + 6.0120e-03, 7.8735e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_img.weight': tensor([[ 0.0079, 0.0034, 0.0430, ..., -0.0327, 0.0042, 0.0146], + [-0.0278, 0.0280, 0.0249, ..., -0.0046, -0.0378, 0.0271], + [ 0.0142, -0.0120, 0.0312, ..., -0.0074, 0.0253, -0.0118], + ..., + [ 0.0245, -0.0152, 0.0027, ..., -0.0452, 0.0216, -0.0209], + [-0.0071, 0.0074, -0.0471, ..., -0.0101, 0.0157, -0.0083], + [ 0.0085, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0028, -0.0122, -0.0115], + [ 0.0137, 0.0062, -0.0188, ..., 0.0157, 0.0082, -0.0047], + [-0.0306, -0.0051, -0.0273, ..., 0.0103, 0.0337, -0.0048], + ..., + [-0.0071, 0.0003, 0.0063, ..., 0.0209, -0.0106, 0.0081], + [ 0.0194, 0.0210, 0.0123, ..., -0.0146, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0109, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img.weight': tensor([[-0.0003, -0.0076, 0.0173, ..., 0.0084, 0.0356, 0.0217], + [-0.0498, -0.0049, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0104, -0.0013, -0.0139, ..., 0.0041, 0.0269, 0.0082], + ..., + [-0.0045, 0.0120, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0077, -0.0162, ..., -0.0059, -0.0076, -0.0121], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0134, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.mlp.layer1.weight': tensor([[-0.0031, -0.0219, -0.0079, ..., 0.0020, -0.0061, -0.0039], + [ 0.0006, 0.0044, 0.0055, ..., 0.0060, -0.0048, -0.0092], + [-0.0221, 0.0100, 0.0028, ..., -0.0045, -0.0090, -0.0073], + ..., + [ 0.0286, 0.0098, 0.0014, ..., 0.0140, 0.0062, 0.0143], + [-0.0159, -0.0058, 0.0041, ..., 0.0157, -0.0101, -0.0102], + [ 0.0052, -0.0114, -0.0042, ..., -0.0054, 0.0070, -0.0181]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.mlp.layer2.weight': tensor([[-0.0096, -0.0122, -0.0084, ..., -0.0148, -0.0040, -0.0079], + [ 0.0123, 0.0056, 0.0131, ..., 0.0037, 0.0097, -0.0077], + [ 0.0052, -0.0040, -0.0037, ..., -0.0010, -0.0015, 0.0038], + ..., + [ 0.0037, -0.0034, 0.0304, ..., -0.0075, -0.0130, 0.0051], + [ 0.0050, 0.0050, 0.0135, ..., 0.0075, 0.0049, -0.0062], + [ 0.0220, -0.0164, -0.0082, ..., 0.0068, 0.0042, 0.0048]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0432, -0.0420, -0.0415, ..., 0.0050, 0.0053, 0.0052], + [ 0.0352, 0.0339, 0.0330, ..., -0.0036, -0.0035, -0.0036], + [-0.0422, -0.0403, -0.0398, ..., 0.0050, 0.0051, 0.0050], + ..., + [-0.0476, -0.0464, -0.0449, ..., 0.0034, 0.0035, 0.0036], + [-0.0413, -0.0410, -0.0393, ..., 0.0036, 0.0039, 0.0039], + [-0.0251, -0.0240, -0.0228, ..., 0.0011, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0042, -0.0036, 0.0035, ..., 0.0033, 0.0033, 0.0034], + [ 0.0006, 0.0030, 0.0020, ..., 0.0028, 0.0019, -0.0021], + [-0.0068, 0.0038, -0.0041, ..., -0.0077, -0.0069, -0.0038], + ..., + [-0.0231, -0.0007, -0.0286, ..., -0.0308, -0.0292, 0.0007], + [-0.0117, -0.0054, -0.0129, ..., -0.0142, -0.0121, 0.0092], + [ 0.0069, -0.0063, 0.0044, ..., 0.0069, 0.0062, -0.0011]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_cross_attn.1.weight': tensor([[ 6.2500e-02, 6.0791e-02, 6.0303e-02, ..., -1.9550e-05, + -4.8876e-05, -4.3213e-06], + [ 7.8125e-03, 7.8125e-03, 7.8125e-03, ..., -3.9577e-05, + -1.1349e-04, -2.2888e-04], + [-2.0264e-02, -1.9531e-02, -1.8555e-02, ..., -6.7139e-04, + -6.7139e-04, -6.8283e-04], + ..., + [-2.6489e-02, -2.6489e-02, -2.5635e-02, ..., -3.9864e-04, + -3.5667e-04, -3.1853e-04], + [-2.7100e-02, -2.6489e-02, -2.5635e-02, ..., -2.0123e-04, + -1.5545e-04, -2.0885e-04], + [ 5.7373e-02, 5.5664e-02, 5.4932e-02, ..., -2.0409e-04, + -8.5831e-05, -2.3174e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0074, -0.0041, -0.0145, ..., 0.0043, 0.0080, -0.0095], + [ 0.0068, 0.0026, -0.0005, ..., -0.0011, -0.0033, 0.0017], + [ 0.0147, -0.0042, -0.0042, ..., 0.0043, -0.0070, 0.0095], + ..., + [ 0.0679, 0.0045, -0.0058, ..., 0.0059, -0.0175, 0.0630], + [ 0.0078, -0.0104, -0.0044, ..., -0.0155, -0.0220, 0.0183], + [-0.0065, 0.0087, -0.0153, ..., 0.0106, 0.0091, -0.0083]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_mlp.1.weight': tensor([[ 0.0096, 0.0090, 0.0082, ..., 0.0003, 0.0002, 0.0002], + [ 0.0325, 0.0317, 0.0315, ..., -0.0037, -0.0037, -0.0036], + [ 0.0457, 0.0444, 0.0439, ..., -0.0038, -0.0038, -0.0038], + ..., + [ 0.0085, 0.0082, 0.0088, ..., -0.0020, -0.0020, -0.0019], + [-0.0181, -0.0167, -0.0159, ..., 0.0015, 0.0014, 0.0014], + [ 0.0052, 0.0052, 0.0052, ..., -0.0024, -0.0025, -0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_mlp.2.weight': tensor([[-0.0024, -0.0014, -0.0033, ..., 0.0035, 0.0031, 0.0024], + [-0.0004, -0.0010, 0.0036, ..., 0.0014, 0.0003, 0.0015], + [ 0.0036, -0.0046, -0.0056, ..., -0.0038, -0.0011, -0.0038], + ..., + [-0.0011, 0.0083, 0.0131, ..., -0.0092, 0.0016, -0.0082], + [ 0.0167, 0.0129, 0.0238, ..., -0.0120, -0.0175, -0.0125], + [ 0.0157, -0.0153, -0.0156, ..., -0.0150, -0.0161, -0.0140]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.after_proj.weight': tensor([[-0.0284, 0.0079, 0.0044, ..., -0.0072, -0.0132, 0.0128], + [ 0.0086, 0.0190, 0.0042, ..., -0.0061, -0.0064, 0.0031], + [ 0.0120, -0.0156, -0.0128, ..., -0.0083, 0.0104, 0.0078], + ..., + [-0.0187, -0.0171, 0.0166, ..., -0.0065, -0.0040, -0.0048], + [-0.0098, -0.0085, 0.0060, ..., 0.0056, 0.0134, -0.0042], + [-0.0182, 0.0089, -0.0125, ..., -0.0096, -0.0066, -0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.after_proj.bias': tensor([ 0.0125, 0.0034, 0.0019, ..., 0.0067, 0.0037, -0.0056], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_proj.weight': tensor([[ 0.0243, -0.0233, -0.0165, ..., 0.0219, -0.0100, -0.0254], + [-0.0001, -0.0242, -0.0025, ..., -0.0129, 0.0074, 0.0037], + [ 0.0148, 0.0154, 0.0070, ..., 0.0044, -0.0095, 0.0009], + ..., + [-0.0036, -0.0067, -0.0034, ..., 0.0040, 0.0106, -0.0075], + [ 0.0281, 0.0144, -0.0022, ..., 0.0077, -0.0160, 0.0281], + [ 0.0074, -0.0193, 0.0089, ..., -0.0330, 0.0123, -0.0231]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_norm.weight': tensor([ 1.5625e+00, 4.9414e-01, 5.1562e-01, 3.0273e-01, 1.4922e+00, + 4.3335e-03, -1.1719e-02, -4.3392e-05, 1.8616e-03, 8.9722e-03, + 1.4827e-06, 3.5623e-08, 1.2266e+00, -6.4392e-03, -3.1876e-17, + -3.2715e-02, -1.3367e-02, -4.3213e-02, 2.4414e-03, 1.8188e-02, + 1.6895e-01, 2.8372e-05, 6.4062e-01, 1.5859e+00, 6.0547e-01, + 6.8750e-01, 4.7656e-01, 1.6699e-01, 2.0020e-02, 5.5847e-03, + 2.7710e-02, -4.9591e-05, 6.1340e-03, -2.0874e-02, 3.3691e-02, + 4.4434e-02, 1.2451e-01, 8.9722e-03, -7.6599e-03, 1.5736e-05, + -5.0545e-05, -2.2507e-04, 1.2146e-02, 1.3125e+00, 1.3184e-01, + 1.3906e+00, 6.6797e-01, 2.1094e-01, -3.7598e-02, -1.5869e-02, + 2.9297e-02, 5.1025e-02, 1.3477e-01, 3.3855e-05, 2.4080e-05, + -5.6252e-07, 4.5061e-05, -7.8678e-05, -6.1279e-02, -4.2783e-09, + 4.6875e-02, -1.9932e-04, -3.3691e-02, -3.5889e-02, 8.7500e-01, + 1.6016e+00, 1.4688e+00, 1.4531e+00, 3.1250e-01, -4.6997e-03, + 5.7068e-03, 3.9339e-05, 1.2939e-02, -4.2114e-03, -1.9932e-04, + 9.0525e-07, 3.2715e-02, 2.4780e-02, 7.6050e-15, 8.1250e-01, + 1.0078e+00, -4.6387e-03, -1.1368e-03, 3.2959e-02, 1.2988e-01, + 2.0508e-02, 1.5234e+00, 5.6641e-01, 1.3984e+00, 1.5000e+00, + 2.8125e-01, 5.1562e-01, 5.1025e-02, -1.1902e-02, 2.1606e-02, + -1.8024e-04, -2.0752e-03, -3.7231e-03, 1.8463e-03, 3.2959e-02, + 1.2207e-01, 2.5177e-04, 1.1641e+00, 1.7405e-05, 2.4915e-05, + -4.7684e-05, -8.5354e-05, 9.6875e-01, 1.5391e+00, 1.4844e+00, + 1.5469e+00, 3.7891e-01, 3.2227e-02, 4.5410e-02, 2.2827e-02, + 1.9287e-02, 1.4746e-01, 6.7711e-05, -7.6890e-06, 1.7658e-06, + 1.8120e-05, 9.2030e-05, 1.1094e+00, 2.7831e-10, 2.4536e-02, + -1.2219e-05, 7.6562e-01, 1.1172e+00], dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.self_attn.k_proj.weight': tensor([[-0.0388, 0.0049, -0.0071, ..., 0.0249, 0.0139, 0.0238], + [-0.0056, 0.0130, -0.0036, ..., -0.0114, 0.0283, -0.0051], + [-0.0294, 0.0200, -0.0027, ..., 0.0057, 0.0228, -0.0034], + ..., + [-0.0057, 0.0018, -0.0045, ..., 0.0026, 0.0011, 0.0127], + [ 0.0050, 0.0120, -0.0009, ..., -0.0046, -0.0048, 0.0089], + [ 0.0190, -0.0225, -0.0019, ..., -0.0131, 0.0058, -0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.k_norm.weight': tensor([ 1.6328e+00, 8.9844e-01, 3.9648e-01, 3.6328e-01, 1.4922e+00, + 4.3030e-03, 1.6235e-02, -4.3392e-05, 6.6528e-03, 8.9722e-03, + 3.5018e-06, -1.6065e-08, 1.2266e+00, -6.0425e-03, -8.1966e-17, + -2.6489e-02, -4.1199e-03, 1.1768e-01, 4.2725e-03, 1.9836e-03, + 2.7466e-02, -7.7724e-05, 6.5625e-01, 1.4297e+00, 2.5391e-01, + 7.0312e-01, 5.5859e-01, 2.3926e-01, -1.8677e-02, 1.3000e-02, + 2.9663e-02, 1.3947e-05, 6.5308e-03, 5.4688e-02, 3.4912e-02, + -1.8311e-02, -1.2573e-02, 8.9722e-03, 2.6489e-02, 1.5974e-05, + -8.2016e-05, -2.2030e-04, 1.2024e-02, 1.5078e+00, 9.9219e-01, + 1.3750e+00, 7.3047e-01, 2.5586e-01, 6.5308e-03, 6.7871e-02, + 3.1738e-02, 7.4158e-03, 3.6774e-03, 6.2466e-05, -1.4484e-05, + -4.8429e-07, 2.9206e-05, 4.9114e-05, 9.8877e-03, -4.5693e-09, + -2.2217e-02, -1.9741e-04, -4.4434e-02, -3.8818e-02, 7.5781e-01, + 1.4609e+00, 1.5000e+00, 1.4375e+00, 2.9688e-01, 4.4556e-03, + -1.6937e-03, 3.6716e-05, 2.1515e-03, 4.3030e-03, -1.9932e-04, + -8.9034e-07, -3.0029e-02, 2.5024e-02, 7.4940e-15, 8.1250e-01, + 1.0078e+00, 5.7617e-02, 1.6022e-03, 1.1719e-02, -3.0884e-02, + 2.0752e-02, 1.4844e+00, 9.2969e-01, 1.5156e+00, 1.5391e+00, + 4.8584e-02, 5.1172e-01, 9.9487e-03, 6.8970e-03, 1.5625e-02, + 2.0885e-04, -6.2180e-04, 4.5166e-02, 4.4250e-03, 1.0498e-02, + 1.1414e-02, -1.1492e-04, 1.1641e+00, 1.7643e-05, 2.4319e-05, + -4.7207e-05, -4.1246e-05, 6.7188e-01, 1.3672e+00, 1.3828e+00, + 1.6641e+00, 3.8086e-01, -4.3701e-02, 1.1963e-01, 2.4170e-02, + 2.2583e-02, -1.5259e-02, -1.0729e-06, -2.0266e-06, -1.4156e-06, + 4.1485e-05, -1.6212e-04, 1.1094e+00, -3.0923e-10, -1.7853e-03, + -8.9407e-06, 7.6562e-01, 1.1172e+00], dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.self_attn.v_proj.weight': tensor([[ 0.0025, -0.0052, -0.0096, ..., -0.0293, -0.0051, 0.0109], + [ 0.0068, 0.0021, -0.0122, ..., -0.0017, -0.0008, -0.0042], + [-0.0245, 0.0041, -0.0022, ..., 0.0148, 0.0055, -0.0143], + ..., + [ 0.0137, 0.0065, -0.0337, ..., -0.0454, 0.0104, -0.0117], + [ 0.0017, -0.0108, 0.0018, ..., -0.0069, -0.0074, -0.0117], + [-0.0015, 0.0028, -0.0188, ..., -0.0042, 0.0068, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.output_proj.weight': tensor([[-0.0040, 0.0315, -0.0271, ..., -0.0048, -0.0188, 0.0134], + [ 0.0038, -0.0175, -0.0282, ..., 0.0054, 0.0064, -0.0032], + [-0.0034, -0.0038, 0.0041, ..., 0.0038, -0.0100, 0.0074], + ..., + [ 0.0094, -0.0041, -0.0128, ..., 0.0277, 0.0123, -0.0049], + [ 0.0029, 0.0190, -0.0044, ..., 0.0171, 0.0071, -0.0211], + [-0.0037, -0.0123, 0.0038, ..., -0.0239, 0.0061, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_proj.weight': tensor([[ 0.0055, -0.0037, 0.0042, ..., -0.0042, 0.0129, 0.0038], + [ 0.0061, 0.0057, 0.0023, ..., 0.0026, 0.0038, -0.0031], + [ 0.0036, -0.0056, 0.0003, ..., -0.0069, 0.0186, -0.0042], + ..., + [-0.0025, 0.0046, 0.0038, ..., 0.0108, 0.0037, 0.0046], + [-0.0059, -0.0049, 0.0035, ..., 0.0124, 0.0095, 0.0079], + [ 0.0080, 0.0013, -0.0031, ..., -0.0109, 0.0103, -0.0040]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_norm.weight': tensor([ 6.7749e-03, 1.2305e-01, 2.2339e-02, 6.0156e-01, 2.3315e-02, + 9.6436e-03, 2.0386e-02, 4.0588e-03, 1.5137e-02, 1.0645e-01, + 2.2583e-02, 4.7852e-02, 2.6172e-01, 2.6489e-02, 1.7969e-01, + 1.8433e-02, 4.6875e-02, 2.4048e-02, 5.5847e-03, 8.3008e-03, + 1.2268e-02, 1.5918e-01, 2.7148e-01, 1.9238e-01, 2.1118e-02, + 1.4941e-01, 4.4922e-02, 8.5449e-03, 4.0820e-01, 1.3611e-02, + 6.3782e-03, 5.0391e-01, 4.4922e-02, 4.1748e-02, 2.0752e-02, + 8.9722e-03, 7.5684e-02, 5.2344e-01, 5.9375e-01, 3.5400e-03, + 1.3062e-02, 1.0156e-01, 8.3618e-03, 5.5859e-01, 1.6479e-02, + 3.2715e-02, 8.4839e-03, 4.2773e-01, 4.2419e-03, 2.1484e-02, + 6.3477e-02, 2.2949e-02, 1.7334e-02, 5.6641e-01, -7.0095e-05, + 3.2227e-02, 3.3691e-02, 6.7383e-02, 1.5198e-02, 1.5198e-02, + 2.5330e-03, 1.2329e-02, 1.5717e-03, 3.5547e-01, 1.4648e-02, + -1.2684e-04, 1.8433e-02, 5.8350e-02, 7.4707e-02, 9.2163e-03, + 9.2773e-03, 6.3672e-01, 1.9789e-05, 6.0938e-01, 1.5381e-02, + 6.7139e-04, 9.5825e-03, 2.9297e-01, 2.9907e-02, 2.9564e-04, + 9.3262e-02, 8.3008e-03, 4.1748e-02, 7.6294e-03, 2.4170e-02, + 5.2490e-03, 2.2827e-02, 6.4941e-02, 2.0898e-01, 9.8877e-03, + 5.9814e-03, 1.3489e-02, 4.0527e-02, 2.0752e-02, 3.5156e-02, + 2.5269e-02, 9.3384e-03, 2.4414e-02, 5.0293e-02, 1.9531e-02, + 2.9907e-02, 1.5625e-02, -3.8300e-03, 5.1514e-02, 6.1719e-01, + 5.2734e-02, 1.1749e-03, 4.9219e-01, 1.8799e-02, 2.3315e-02, + 1.1902e-02, 2.0294e-03, -6.1512e-05, 8.0078e-02, 4.6094e-01, + 1.4099e-02, 2.3682e-02, 6.1328e-01, 1.1084e-01, 3.0518e-02, + 1.1169e-02, 6.9922e-01, 1.4453e-01, 1.3916e-02, 7.8125e-03, + 2.4414e-03, 1.8921e-02, 1.1520e-03], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_proj.weight': tensor([[ 0.0038, -0.0034, 0.0050, ..., -0.0018, -0.0112, 0.0114], + [ 0.0039, 0.0003, -0.0025, ..., 0.0014, -0.0126, -0.0041], + [ 0.0024, -0.0016, -0.0035, ..., 0.0120, -0.0132, -0.0042], + ..., + [ 0.0052, 0.0046, -0.0034, ..., 0.0023, 0.0061, -0.0125], + [ 0.0093, 0.0029, -0.0020, ..., -0.0042, 0.0042, 0.0027], + [-0.0091, 0.0007, -0.0101, ..., -0.0042, -0.0160, 0.0046]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_norm.weight': tensor([ 6.7749e-03, 1.2305e-01, 2.2461e-02, 6.0156e-01, 2.3193e-02, + 9.4604e-03, 2.0386e-02, 4.0588e-03, 1.5381e-02, 1.0645e-01, + 2.2583e-02, 4.6143e-02, 2.6172e-01, 2.6611e-02, 1.7969e-01, + 1.9409e-02, 4.6387e-02, 2.4048e-02, 5.5237e-03, 8.1787e-03, + 1.2390e-02, 1.6016e-01, 2.7344e-01, 1.9238e-01, 2.0752e-02, + 1.4941e-01, 4.3945e-02, 6.3477e-03, 4.1016e-01, 1.3489e-02, + 6.6528e-03, 5.0391e-01, 4.4922e-02, 4.1992e-02, 2.1118e-02, + 8.3008e-03, 7.4707e-02, 5.2344e-01, 5.9375e-01, 3.6011e-03, + 1.3062e-02, 1.0107e-01, 8.3618e-03, 5.5859e-01, 1.6846e-02, + 3.2959e-02, 8.7891e-03, 4.2773e-01, 4.3335e-03, 2.1362e-02, + 6.3477e-02, 2.3193e-02, 1.7334e-02, 5.6641e-01, -7.0095e-05, + 3.2227e-02, 3.3203e-02, 6.7383e-02, 1.5259e-02, 1.3672e-02, + 2.4261e-03, 1.2268e-02, 2.3956e-03, 3.5547e-01, 1.4221e-02, + 2.9755e-04, 1.7822e-02, 5.8350e-02, 7.4707e-02, 9.3384e-03, + 9.5825e-03, 6.3672e-01, 1.9789e-05, 6.0938e-01, 1.5625e-02, + -5.8365e-04, 9.6436e-03, 2.9297e-01, 3.0640e-02, 3.8910e-04, + 9.3750e-02, 8.1177e-03, 4.0771e-02, 7.3547e-03, 2.3682e-02, + 5.2490e-03, 2.3193e-02, 6.5430e-02, 2.0898e-01, 9.8877e-03, + 6.0120e-03, 1.2329e-02, 4.0527e-02, 2.0996e-02, 3.6133e-02, + 2.5269e-02, 9.3994e-03, 2.4414e-02, 5.0293e-02, 1.9409e-02, + 2.9175e-02, 1.5747e-02, 3.8300e-03, 5.2246e-02, 6.1719e-01, + 5.2734e-02, 1.1444e-03, 4.9219e-01, 1.8433e-02, 2.3071e-02, + 1.2695e-02, 2.0294e-03, -5.4836e-05, 8.0078e-02, 4.6094e-01, + 1.3672e-02, 2.3438e-02, 6.1328e-01, 1.1182e-01, 3.0640e-02, + 1.0864e-02, 7.0312e-01, 1.4355e-01, 1.2756e-02, 7.5378e-03, + 2.3346e-03, 1.8677e-02, -9.3460e-04], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.v_proj.weight': tensor([[ 0.0133, -0.0066, -0.0176, ..., -0.0031, -0.0162, -0.0116], + [-0.0157, 0.0014, -0.0052, ..., -0.0026, -0.0210, 0.0210], + [-0.0116, -0.0044, -0.0012, ..., -0.0032, 0.0299, 0.0249], + ..., + [-0.0052, 0.0014, -0.0125, ..., 0.0035, 0.0112, -0.0006], + [ 0.0284, 0.0068, 0.0254, ..., 0.0093, 0.0096, 0.0153], + [ 0.0162, 0.0038, 0.0074, ..., 0.0071, -0.0186, -0.0244]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.output_proj.weight': tensor([[ 0.0153, 0.0042, -0.0161, ..., 0.0133, 0.0081, 0.0150], + [ 0.0088, -0.0010, -0.0029, ..., 0.0084, 0.0030, 0.0031], + [-0.0039, -0.0069, 0.0044, ..., -0.0048, 0.0067, -0.0067], + ..., + [ 0.0047, -0.0082, 0.0043, ..., 0.0047, -0.0096, 0.0056], + [ 0.0045, -0.0133, 0.0067, ..., -0.0403, -0.0249, -0.0233], + [-0.0021, -0.0025, 0.0154, ..., 0.0086, 0.0036, 0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0068, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [ 0.0038, 0.0150, -0.0159, ..., -0.0034, -0.0056, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0068, -0.0057], + [-0.0337, -0.0031, 0.0048, ..., -0.0032, 0.0172, 0.0231], + [-0.0255, 0.0004, -0.0044, ..., 0.0408, 0.0430, -0.0120]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0033], + [ 0.0198, -0.0208, 0.0129, ..., 0.0089, -0.0337, -0.0051], + [ 0.0302, 0.0101, -0.0018, ..., -0.0304, 0.0199, -0.0039], + ..., + [ 0.0171, -0.0276, -0.0066, ..., 0.0231, 0.0182, 0.0011], + [ 0.0221, 0.0388, 0.0047, ..., 0.0042, -0.0488, -0.0361], + [ 0.0045, 0.0118, -0.0059, ..., 0.0010, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img.weight': tensor([[ 0.0273, -0.0041, 0.0118, ..., 0.0022, 0.0347, 0.0615], + [-0.0258, -0.0047, -0.0197, ..., 0.0139, -0.0042, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0097, 0.0132, 0.0184, ..., 0.0110, 0.0089, 0.0515], + [-0.0017, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.mlp.layer1.weight': tensor([[-0.0108, 0.0063, 0.0072, ..., -0.0044, -0.0053, -0.0164], + [-0.0096, 0.0017, -0.0006, ..., -0.0088, 0.0035, -0.0081], + [-0.0038, 0.0062, 0.0038, ..., -0.0022, -0.0123, -0.0510], + ..., + [-0.0118, 0.0111, 0.0038, ..., 0.0068, -0.0043, -0.0075], + [-0.0062, 0.0075, 0.0042, ..., -0.0066, -0.0088, -0.0383], + [-0.0030, 0.0275, -0.0038, ..., 0.0016, 0.0046, -0.0094]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.mlp.layer2.weight': tensor([[-0.0107, -0.0052, -0.0210, ..., 0.0140, -0.0079, -0.0112], + [-0.0051, 0.0047, -0.0121, ..., -0.0032, 0.0078, -0.0229], + [-0.0051, 0.0137, -0.0172, ..., 0.0005, 0.0127, -0.0106], + ..., + [-0.0009, 0.0081, -0.0381, ..., 0.0082, 0.0142, -0.0155], + [-0.0127, 0.0184, 0.0122, ..., 0.0067, 0.0096, -0.0037], + [-0.0101, -0.0040, -0.0364, ..., -0.0003, 0.0109, -0.0334]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0289, 0.0291, 0.0275, ..., -0.0041, -0.0042, -0.0043], + [-0.0254, -0.0244, -0.0242, ..., 0.0043, 0.0047, 0.0046], + [-0.0297, -0.0293, -0.0286, ..., 0.0045, 0.0048, 0.0048], + ..., + [-0.0322, -0.0312, -0.0305, ..., 0.0021, 0.0022, 0.0023], + [-0.0327, -0.0325, -0.0317, ..., 0.0020, 0.0021, 0.0021], + [-0.0046, -0.0046, -0.0045, ..., 0.0001, 0.0001, 0.0001]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0049, -0.0042, -0.0014, ..., -0.0048, -0.0051, -0.0014], + [ 0.0017, -0.0020, -0.0017, ..., -0.0018, -0.0022, 0.0031], + [ 0.0099, -0.0096, -0.0102, ..., -0.0041, -0.0034, 0.0007], + ..., + [ 0.0388, -0.0427, -0.0432, ..., -0.0249, -0.0217, 0.0142], + [-0.0041, 0.0042, -0.0049, ..., 0.0044, 0.0045, 0.0060], + [-0.0148, 0.0173, 0.0148, ..., 0.0162, 0.0146, 0.0165]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_cross_attn.1.weight': tensor([[ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., -1.3065e-04, + -6.0320e-05, -2.1267e-04], + [ 6.1768e-02, 6.0303e-02, 5.8594e-02, ..., 3.3569e-04, + 4.4632e-04, 4.7874e-04], + [-5.3955e-02, -5.2490e-02, -5.1025e-02, ..., -5.2154e-06, + 5.0545e-05, -5.7459e-05], + ..., + [ 3.3447e-02, 3.2715e-02, 3.1982e-02, ..., 1.0204e-04, + 3.3617e-05, 6.2943e-05], + [ 5.9570e-02, 5.8105e-02, 5.7617e-02, ..., 5.2261e-04, + 5.6458e-04, 5.5695e-04], + [-4.8828e-02, -4.8096e-02, -4.7363e-02, ..., -3.2997e-04, + -3.7384e-04, -3.3379e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0254, 0.0063, -0.0156, ..., 0.0030, -0.0043, -0.0016], + [-0.0216, 0.0137, 0.0175, ..., -0.0090, -0.0137, 0.0099], + [ 0.0098, 0.0008, -0.0087, ..., 0.0060, 0.0008, -0.0008], + ..., + [ 0.0225, 0.0156, -0.0128, ..., 0.0072, 0.0080, 0.0045], + [ 0.0117, 0.0444, -0.0057, ..., 0.0040, 0.0225, -0.0079], + [-0.0187, 0.0161, 0.0192, ..., -0.0135, -0.0082, 0.0210]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_mlp.1.weight': tensor([[-0.0571, -0.0557, -0.0542, ..., 0.0029, 0.0031, 0.0031], + [-0.0259, -0.0244, -0.0232, ..., 0.0008, 0.0008, 0.0007], + [ 0.0344, 0.0332, 0.0332, ..., -0.0016, -0.0016, -0.0016], + ..., + [ 0.0304, 0.0294, 0.0295, ..., -0.0018, -0.0017, -0.0015], + [ 0.0356, 0.0342, 0.0327, ..., -0.0010, -0.0009, -0.0010], + [-0.0295, -0.0288, -0.0282, ..., 0.0021, 0.0021, 0.0020]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_mlp.2.weight': tensor([[-0.0233, -0.0145, 0.0118, ..., 0.0041, 0.0162, -0.0124], + [-0.0010, -0.0058, -0.0026, ..., -0.0016, 0.0057, 0.0014], + [-0.0004, -0.0009, 0.0005, ..., -0.0008, 0.0005, 0.0001], + ..., + [ 0.0432, -0.0041, -0.0190, ..., -0.0137, 0.0040, 0.0197], + [ 0.0226, -0.0126, -0.0173, ..., -0.0151, 0.0124, 0.0167], + [ 0.0072, -0.0306, -0.0190, ..., -0.0220, 0.0315, 0.0182]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.after_proj.weight': tensor([[-0.0034, 0.0076, -0.0066, ..., -0.0079, 0.0073, 0.0118], + [ 0.0095, -0.0077, 0.0047, ..., -0.0018, 0.0079, 0.0058], + [ 0.0027, 0.0077, 0.0167, ..., 0.0063, 0.0069, 0.0176], + ..., + [ 0.0086, -0.0040, 0.0145, ..., 0.0044, -0.0072, 0.0212], + [ 0.0033, -0.0074, 0.0148, ..., -0.0079, -0.0046, -0.0078], + [-0.0034, 0.0047, -0.0098, ..., -0.0081, 0.0216, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.after_proj.bias': tensor([-0.0038, -0.0012, 0.0049, ..., -0.0038, 0.0101, -0.0013], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_proj.weight': tensor([[-0.0121, 0.0339, 0.0119, ..., -0.0087, 0.0035, -0.0254], + [-0.0052, 0.0046, 0.0090, ..., 0.0026, 0.0165, -0.0212], + [ 0.0130, 0.0161, 0.0101, ..., -0.0052, -0.0282, 0.0151], + ..., + [-0.0036, -0.0046, 0.0041, ..., 0.0019, 0.0104, 0.0014], + [-0.0120, -0.0302, 0.0251, ..., -0.0388, 0.0052, -0.0098], + [-0.0039, -0.0037, 0.0004, ..., 0.0075, 0.0177, 0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_norm.weight': tensor([ 9.3750e-01, 1.7422e+00, 1.4844e+00, 1.5547e+00, 1.5430e-01, + 1.2207e-01, 1.7212e-02, -1.6065e-08, -2.3954e-33, 3.5553e-03, + 1.1353e-02, 1.0791e-01, -5.8594e-03, -4.5166e-03, 1.6022e-03, + -1.0490e-04, 6.5430e-02, 1.0391e+00, 1.1094e+00, 7.7820e-04, + 4.1485e-05, 1.2891e+00, 2.0312e+00, 1.8516e+00, 1.0625e+00, + 8.8281e-01, 1.3750e+00, -1.2817e-02, -6.9046e-04, 3.4714e-04, + -2.3041e-03, -2.5940e-04, 3.7354e-02, 1.1641e+00, 6.1035e-03, + 1.1172e+00, 1.6113e-02, 3.1710e-05, 1.4221e-02, -7.5378e-03, + 2.0862e-05, 1.6212e-05, 1.5020e-05, 2.0938e+00, 1.3516e+00, + 1.4922e+00, 1.5234e+00, 1.4688e+00, 9.4238e-02, 5.4932e-04, + 3.1982e-02, 2.7148e-01, 1.1719e+00, 1.1475e-01, 2.1362e-03, + 8.2520e-02, 2.8839e-03, 1.0107e-01, -4.4107e-05, 3.6163e-03, + -1.1978e-03, 5.2185e-03, -5.2490e-03, 1.1719e+00, 1.9375e+00, + 6.9531e-01, 8.5547e-01, 1.3828e+00, 4.5410e-02, -2.3193e-02, + 1.1475e-02, 3.3993e-08, -3.8037e-33, 9.4604e-03, 1.8921e-03, + -3.1586e-03, -1.5450e-04, 2.0386e-02, -4.1748e-02, 2.5635e-03, + 4.6387e-02, 4.7607e-03, 1.8188e-02, 8.7738e-04, 9.6321e-05, + -5.2185e-03, 8.4375e-01, 8.2031e-01, 1.5312e+00, 1.4609e+00, + 7.6562e-01, -5.9814e-03, -2.0885e-04, -5.9605e-05, -2.3651e-03, + 2.6894e-04, 2.1973e-02, 1.3770e-01, 2.3804e-03, 4.5654e-02, + -2.1362e-02, -8.4639e-06, 1.2891e-01, 6.4850e-05, 5.7983e-03, + -1.5736e-05, -2.9206e-05, 5.1953e-01, 1.7266e+00, 1.4922e+00, + 8.3984e-01, 7.5781e-01, 9.4727e-02, 4.0054e-04, 2.5391e-02, + 1.1562e+00, 8.0078e-02, 1.1875e+00, -7.3910e-05, 9.0332e-03, + 1.9043e-02, 4.7852e-02, 2.4438e-05, -2.1210e-03, 2.1648e-04, + -2.1729e-02, 1.1328e+00, -9.9487e-03], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.self_attn.k_proj.weight': tensor([[-1.5747e-02, 6.3171e-03, 2.8381e-03, ..., -1.1230e-02, + -3.0212e-03, 3.7842e-03], + [-1.0864e-02, -2.5635e-02, 1.5503e-02, ..., -7.3242e-03, + -4.9438e-03, -5.3406e-03], + [ 5.0964e-03, 4.3945e-03, 9.0942e-03, ..., 1.7090e-02, + -6.8665e-03, 3.7384e-03], + ..., + [-4.6082e-03, 4.9133e-03, 4.7913e-03, ..., 1.9836e-03, + -3.8605e-03, 3.3417e-03], + [-1.3550e-02, -1.5736e-05, 2.5757e-02, ..., -1.5381e-02, + 7.1716e-03, 5.3406e-03], + [ 1.9226e-03, 4.5166e-03, -7.6904e-03, ..., -4.6692e-03, + 6.8665e-03, 2.1362e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.k_norm.weight': tensor([ 8.7891e-01, 1.7109e+00, 1.5000e+00, 1.4922e+00, 1.5625e-01, + 1.2793e-01, 1.1902e-02, -6.7521e-09, -1.7935e-33, -1.6251e-03, + 1.1169e-02, 1.0889e-01, -1.0925e-02, 3.1006e-02, -2.0142e-03, + 9.8705e-05, 1.3123e-02, 1.0391e+00, 1.1094e+00, 7.6675e-04, + 7.2479e-05, 1.2891e+00, 1.7969e+00, 1.7188e+00, 1.2891e-01, + 3.8281e-01, 1.4219e+00, -1.5015e-02, 3.2425e-04, -3.0136e-04, + -2.4567e-03, 2.9564e-04, -3.6133e-02, 1.1641e+00, 2.5787e-03, + 1.1094e+00, 1.6113e-02, -3.0518e-05, -1.4954e-02, -7.6599e-03, + -1.0788e-05, 2.3365e-05, -5.0306e-05, 1.6016e+00, 6.2109e-01, + 1.4531e+00, 1.5781e+00, 1.4922e+00, -5.6885e-02, 5.0735e-04, + 1.9653e-02, -9.3994e-03, 1.1719e+00, 1.2354e-01, 2.1515e-03, + 8.2520e-02, -3.9368e-03, 1.0156e-01, -3.3855e-05, 7.0801e-02, + -1.2054e-03, 2.3315e-02, -1.5869e-02, 1.1719e+00, 1.9609e+00, + 9.9219e-01, 7.5781e-01, 1.3672e+00, 2.6855e-02, -1.0193e-02, + -2.4170e-02, 9.8255e-08, 4.4778e-33, 1.0010e-02, 1.4587e-02, + 4.3945e-02, 1.6403e-04, 3.7842e-02, -5.2490e-02, -2.6093e-03, + -4.6692e-03, 1.9287e-02, 1.5198e-02, 8.6975e-04, 9.9182e-05, + 3.2043e-03, 1.3828e+00, 1.3281e+00, 1.6250e+00, 1.5156e+00, + 3.1445e-01, 2.5177e-03, -6.5231e-04, -1.4687e-04, 2.4109e-03, + 3.3951e-04, 4.9744e-03, -2.4170e-02, -1.9684e-03, 7.1411e-03, + 2.1484e-02, -3.4332e-05, 1.2793e-01, -2.4080e-05, 5.7678e-03, + 1.9908e-05, -2.6584e-05, 1.4219e+00, 1.9531e+00, 1.4609e+00, + 1.3086e-01, 7.0703e-01, 5.0659e-03, 4.2915e-04, 4.6082e-03, + 1.1797e+00, -4.9438e-03, 1.1875e+00, 5.9366e-05, -1.8555e-02, + 1.9409e-02, 4.6387e-02, 3.0398e-05, 4.4189e-02, 2.1935e-04, + 2.8198e-02, 1.1328e+00, -1.0071e-02], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.self_attn.v_proj.weight': tensor([[-0.0090, -0.0089, 0.0096, ..., 0.0508, -0.0099, -0.0027], + [-0.0032, 0.0112, -0.0091, ..., 0.0052, -0.0070, -0.0080], + [-0.0040, 0.0182, -0.0139, ..., -0.0025, 0.0050, 0.0128], + ..., + [ 0.0008, 0.0008, -0.0097, ..., -0.0038, 0.0085, 0.0115], + [ 0.0040, 0.0105, -0.0109, ..., 0.0032, -0.0037, -0.0028], + [-0.0021, -0.0089, 0.0081, ..., 0.0151, 0.0078, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.output_proj.weight': tensor([[ 0.0198, 0.0126, 0.0085, ..., 0.0065, 0.0132, -0.0049], + [ 0.0198, -0.0013, 0.0054, ..., -0.0305, -0.0049, 0.0111], + [-0.0098, 0.0184, -0.0212, ..., -0.0133, -0.0063, 0.0136], + ..., + [-0.0292, -0.0199, 0.0059, ..., -0.0051, -0.0112, 0.0065], + [-0.0072, 0.0178, -0.0261, ..., -0.0131, 0.0439, 0.0035], + [-0.0305, -0.0042, 0.0276, ..., 0.0226, -0.0085, -0.0272]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_proj.weight': tensor([[-0.0057, -0.0057, 0.0170, ..., -0.0219, -0.0225, 0.0112], + [-0.0082, 0.0065, 0.0289, ..., -0.0229, -0.0029, 0.0105], + [ 0.0126, -0.0037, -0.0076, ..., 0.0052, 0.0110, -0.0043], + ..., + [ 0.0104, 0.0056, -0.0078, ..., -0.0189, -0.0057, 0.0044], + [ 0.0064, 0.0035, -0.0112, ..., 0.0045, -0.0034, 0.0016], + [-0.0064, -0.0037, 0.0124, ..., 0.0123, -0.0010, 0.0044]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_norm.weight': tensor([0.0967, 0.0156, 0.4414, 0.0576, 0.1416, 0.0320, 0.1377, 0.1201, 0.5195, + 0.0297, 0.0488, 0.0168, 0.0378, 0.0713, 0.0508, 0.1836, 0.4746, 0.0219, + 0.7578, 0.0405, 0.0283, 0.1533, 0.0262, 0.0757, 0.0425, 0.3906, 0.5156, + 0.2617, 0.0684, 0.0405, 0.0231, 0.2178, 0.6797, 0.2402, 0.0322, 0.0349, + 0.0141, 0.0928, 0.0175, 0.0359, 0.0491, 0.0986, 0.0206, 0.0298, 0.2041, + 0.0337, 0.3828, 0.0153, 0.0413, 0.0170, 0.0173, 0.4180, 0.4668, 0.3418, + 0.0708, 0.6211, 0.5977, 0.0447, 0.1060, 0.0198, 0.0767, 0.4648, 0.0071, + 0.5000, 0.1475, 0.3926, 0.0815, 0.0471, 0.0630, 0.4473, 0.0280, 0.4707, + 0.0135, 0.0366, 0.1377, 0.0698, 0.0148, 0.0991, 0.6523, 0.0337, 0.0732, + 0.0908, 0.0388, 0.0503, 0.4492, 0.0601, 0.0640, 0.0879, 0.0189, 0.1396, + 0.0327, 0.1758, 0.0408, 0.0549, 0.0376, 0.0996, 0.0088, 0.3340, 0.5195, + 0.0344, 0.0520, 0.2949, 0.0322, 0.1108, 0.0325, 0.2158, 0.0123, 0.6406, + 0.6211, 0.0175, 0.0161, 0.7188, 0.0410, 0.0160, 0.4805, 0.2090, 0.2695, + 0.6992, 0.0332, 0.0762, 0.0664, 0.5586, 0.6133, 0.0234, 0.0225, 0.1855, + 0.0366, 0.0369], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_proj.weight': tensor([[-3.1586e-03, 1.8158e-03, 4.1504e-03, ..., -1.4114e-03, + 1.0681e-02, 3.1281e-03], + [ 2.8076e-03, -6.7711e-05, -1.6861e-03, ..., 4.7913e-03, + -1.1108e-02, 5.2795e-03], + [-1.9043e-02, 7.6904e-03, -3.2043e-04, ..., 1.8433e-02, + 1.0681e-02, 3.5553e-03], + ..., + [ 5.8594e-03, 3.1128e-03, -5.7983e-03, ..., 9.4986e-04, + -1.9287e-02, 5.4932e-04], + [-9.5825e-03, -1.8539e-03, -1.0681e-02, ..., 6.0425e-03, + -1.0071e-02, 5.9509e-03], + [ 2.0264e-02, -1.4038e-03, 1.0681e-02, ..., -8.6670e-03, + 7.0190e-03, -6.2866e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_norm.weight': tensor([0.0967, 0.0157, 0.4414, 0.0574, 0.1416, 0.0327, 0.1387, 0.1206, 0.5195, + 0.0302, 0.0498, 0.0168, 0.0378, 0.0713, 0.0508, 0.1855, 0.4746, 0.0217, + 0.7539, 0.0403, 0.0281, 0.1533, 0.0264, 0.0752, 0.0422, 0.3867, 0.5156, + 0.2598, 0.0688, 0.0408, 0.0238, 0.2197, 0.6797, 0.2402, 0.0330, 0.0349, + 0.0137, 0.0928, 0.0175, 0.0354, 0.0483, 0.0996, 0.0210, 0.0303, 0.2051, + 0.0337, 0.3828, 0.0154, 0.0410, 0.0166, 0.0179, 0.4180, 0.4668, 0.3418, + 0.0708, 0.6211, 0.6016, 0.0439, 0.1050, 0.0201, 0.0767, 0.4648, 0.0075, + 0.5000, 0.1465, 0.3926, 0.0815, 0.0471, 0.0630, 0.4473, 0.0282, 0.4707, + 0.0132, 0.0366, 0.1377, 0.0693, 0.0148, 0.0996, 0.6523, 0.0334, 0.0723, + 0.0908, 0.0391, 0.0503, 0.4492, 0.0603, 0.0630, 0.0874, 0.0194, 0.1416, + 0.0337, 0.1748, 0.0408, 0.0552, 0.0378, 0.1006, 0.0087, 0.3340, 0.5195, + 0.0337, 0.0525, 0.2949, 0.0325, 0.1108, 0.0325, 0.2158, 0.0122, 0.6406, + 0.6211, 0.0172, 0.0162, 0.7188, 0.0408, 0.0161, 0.4824, 0.2090, 0.2695, + 0.6953, 0.0327, 0.0747, 0.0664, 0.5586, 0.6133, 0.0233, 0.0225, 0.1855, + 0.0378, 0.0374], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.v_proj.weight': tensor([[ 0.0226, 0.0038, -0.0243, ..., 0.0069, -0.0026, -0.0189], + [-0.0203, 0.0140, 0.0005, ..., 0.0066, -0.0025, -0.0311], + [ 0.0251, -0.0085, -0.0001, ..., -0.0010, -0.0099, -0.0452], + ..., + [ 0.0023, -0.0017, 0.0232, ..., -0.0073, -0.0212, 0.0034], + [-0.0327, -0.0032, -0.0256, ..., -0.0060, 0.0122, -0.0121], + [-0.0164, 0.0014, 0.0045, ..., -0.0046, -0.0042, 0.0089]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.output_proj.weight': tensor([[ 0.0017, 0.0038, -0.0017, ..., -0.0120, 0.0074, 0.0096], + [-0.0315, -0.0042, -0.0132, ..., -0.0078, -0.0199, -0.0077], + [ 0.0161, -0.0069, 0.0054, ..., -0.0117, 0.0118, -0.0172], + ..., + [ 0.0181, -0.0146, 0.0189, ..., 0.0119, -0.0060, -0.0081], + [-0.0137, -0.0216, 0.0060, ..., 0.0129, -0.0192, 0.0261], + [ 0.0005, -0.0215, 0.0315, ..., -0.0364, 0.0332, 0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0057, -0.0065, ..., 0.0266, -0.0278, 0.0075], + ..., + [-0.0217, -0.0410, 0.0488, ..., -0.0039, -0.0003, -0.0255], + [ 0.0036, 0.0066, 0.0013, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0059, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0101, 0.0038, ..., -0.0236, -0.0109, -0.0248], + [-0.0079, -0.0013, 0.0293, ..., 0.0072, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0143, 0.0126, -0.0094], + ..., + [-0.0035, 0.0320, -0.0233, ..., -0.0060, -0.0015, 0.0188], + [ 0.0136, -0.0352, 0.0267, ..., -0.0305, -0.0045, 0.0037], + [-0.0205, 0.0140, 0.0025, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img.weight': tensor([[-0.0079, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0029, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0057, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0018], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, -0.0027, -0.0020], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0023, ..., -0.0103, 0.0074, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.mlp.layer1.weight': tensor([[ 0.0151, -0.0038, 0.0084, ..., 0.0068, 0.0133, -0.0104], + [-0.0098, 0.0107, 0.0242, ..., 0.0029, 0.0115, -0.0012], + [-0.0037, 0.0354, -0.0305, ..., 0.0278, 0.0125, 0.0047], + ..., + [-0.0060, 0.0233, 0.0693, ..., -0.0088, 0.0012, -0.0090], + [-0.0055, -0.0082, -0.0050, ..., -0.0095, -0.0016, -0.0164], + [-0.0005, 0.0047, -0.0097, ..., -0.0092, -0.0284, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.mlp.layer2.weight': tensor([[ 0.0067, 0.0117, 0.0032, ..., -0.0220, -0.0221, -0.0071], + [ 0.0104, 0.0247, 0.0068, ..., -0.0092, 0.0356, -0.0181], + [ 0.0040, -0.0050, -0.0037, ..., 0.0150, 0.0275, 0.0042], + ..., + [ 0.0045, 0.0088, 0.0064, ..., -0.0083, -0.0062, 0.0048], + [ 0.0051, -0.0092, -0.0031, ..., 0.0077, 0.0183, 0.0037], + [-0.0036, 0.0048, 0.0170, ..., 0.0052, 0.0349, 0.0073]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0226, -0.0222, -0.0220, ..., 0.0010, 0.0008, 0.0009], + [-0.0315, -0.0306, -0.0298, ..., 0.0026, 0.0027, 0.0027], + [-0.0222, -0.0214, -0.0205, ..., 0.0021, 0.0023, 0.0023], + ..., + [ 0.0605, 0.0588, 0.0569, ..., -0.0037, -0.0037, -0.0037], + [ 0.0496, 0.0479, 0.0464, ..., -0.0032, -0.0031, -0.0030], + [-0.0256, -0.0250, -0.0248, ..., 0.0022, 0.0023, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0061, 0.0044, 0.0062, ..., 0.0099, 0.0130, 0.0059], + [ 0.0005, 0.0014, 0.0015, ..., -0.0015, 0.0004, 0.0016], + [ 0.0015, -0.0013, 0.0005, ..., 0.0037, 0.0038, -0.0014], + ..., + [-0.0237, -0.0762, -0.0615, ..., 0.0369, 0.0354, -0.0693], + [ 0.0089, 0.0082, 0.0085, ..., 0.0182, 0.0150, 0.0092], + [ 0.0164, 0.0099, 0.0124, ..., 0.0187, 0.0208, 0.0113]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_cross_attn.1.weight': tensor([[-8.2031e-02, -8.0078e-02, -7.7637e-02, ..., -1.0443e-04, + -1.2016e-04, -1.0967e-04], + [ 3.0029e-02, 2.9419e-02, 2.8564e-02, ..., 5.6624e-06, + 3.1948e-05, 1.5163e-04], + [ 3.2715e-02, 3.1494e-02, 3.0518e-02, ..., 2.6703e-04, + 2.8992e-04, 1.9836e-04], + ..., + [ 5.5908e-02, 5.4199e-02, 5.3223e-02, ..., -1.8978e-04, + -1.8501e-04, -1.1683e-04], + [ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.8311e-04, + -1.2112e-04, -1.7071e-04], + [-3.0884e-02, -2.9907e-02, -2.9175e-02, ..., -5.8889e-05, + 9.3460e-05, 1.1492e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_cross_attn.2.weight': tensor([[-7.9346e-03, 3.7842e-03, -8.8501e-03, ..., 2.8320e-02, + 3.0396e-02, 3.6926e-03], + [ 6.7139e-03, -9.3994e-03, 1.1841e-02, ..., -2.4605e-04, + 4.5166e-03, 4.6387e-03], + [ 2.0905e-03, -1.2207e-03, 4.6997e-03, ..., -7.9346e-03, + -6.3477e-03, -4.7913e-03], + ..., + [-1.5991e-02, 4.6387e-03, 3.5400e-02, ..., 2.5146e-02, + 3.2227e-02, -8.7891e-03], + [-2.3315e-02, 1.2085e-02, 3.6865e-02, ..., 1.0254e-02, + 1.5869e-02, -1.4099e-02], + [ 1.3977e-02, -8.5449e-03, -7.0801e-03, ..., 1.5717e-03, + 1.6098e-03, -4.5300e-05]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_mlp.1.weight': tensor([[-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., 1.3638e-04, + 2.5940e-04, 2.0313e-04], + [ 1.3245e-02, 1.1841e-02, 1.1597e-02, ..., 3.7193e-04, + 4.5204e-04, 4.4441e-04], + [-7.3242e-02, -7.1777e-02, -7.0801e-02, ..., -4.4823e-04, + -3.7766e-04, -4.1199e-04], + ..., + [-6.5430e-02, -6.4453e-02, -6.3477e-02, ..., 1.6785e-04, + 2.7084e-04, 1.6785e-04], + [-4.4922e-02, -4.4434e-02, -4.4189e-02, ..., 2.0981e-04, + 6.8188e-05, 7.9632e-05], + [ 6.2012e-02, 6.1035e-02, 6.0791e-02, ..., -1.4210e-04, + -1.6975e-04, -1.3351e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_mlp.2.weight': tensor([[-0.0141, 0.0030, -0.0100, ..., -0.0164, -0.0199, 0.0178], + [-0.0017, -0.0063, 0.0064, ..., -0.0015, 0.0049, -0.0028], + [-0.0016, -0.0029, -0.0036, ..., -0.0010, -0.0017, 0.0016], + ..., + [-0.0493, -0.0312, -0.0461, ..., -0.0361, -0.0109, 0.0266], + [ 0.0075, -0.0177, 0.0260, ..., 0.0075, 0.0052, -0.0075], + [ 0.0135, 0.0118, 0.0070, ..., 0.0121, 0.0072, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.after_proj.weight': tensor([[ 2.8229e-03, 1.6724e-02, -1.8311e-02, ..., 7.8125e-03, + 3.9368e-03, -6.8970e-03], + [-8.4229e-03, -3.6316e-03, 7.6599e-03, ..., -9.5844e-05, + 1.6235e-02, -8.6060e-03], + [ 9.6436e-03, 4.3335e-03, -3.9978e-03, ..., 3.8452e-03, + 2.1820e-03, -7.6599e-03], + ..., + [ 1.7456e-02, 1.2329e-02, -5.0354e-03, ..., 7.7515e-03, + -6.1035e-03, 5.2795e-03], + [-6.2866e-03, 7.4463e-03, -3.2349e-03, ..., 1.0437e-02, + -3.9673e-03, 4.2419e-03], + [ 3.4332e-03, -1.4832e-02, -5.6458e-03, ..., 2.1362e-02, + -1.2268e-02, 8.9722e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.after_proj.bias': tensor([ 0.0085, -0.0193, -0.0089, ..., -0.0041, -0.0183, 0.0126], + dtype=torch.bfloat16)} +[06-26 07:35:43|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:43|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 07:35:44|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 2 checkpoint. +[06-26 07:35:44|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/seg(5136ef49-6d8d-42e8-8abf-7dac722a304a) +[06-26 07:35:44|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 3 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/23057a4167b89de89a4a397fdbf3887994d115eb/general/seg/5136ef49-6d8d-42e8-8abf-7dac722a304a_ema_bf16.pt +[06-26 07:35:44|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0143, 0.0022, 0.0062], + [-0.0022, 0.0054, -0.0327, ..., -0.0080, 0.0010, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0134, -0.0026, 0.0283], + ..., + [-0.0026, 0.0125, 0.0248, ..., 0.0437, -0.0073, -0.0192], + [ 0.0277, 0.0148, 0.0054, ..., -0.0015, -0.0091, 0.0146], + [ 0.0437, 0.0366, 0.0082, ..., 0.0016, -0.0129, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0006, -0.0310, 0.0300, ..., -0.0067, -0.0043, -0.0339], + [-0.0104, -0.0066, 0.0111, ..., -0.0004, 0.0092, 0.0520], + [-0.0063, 0.0132, 0.0398, ..., -0.0101, -0.0179, -0.0064], + ..., + [-0.0034, 0.0175, -0.0096, ..., 0.0225, 0.0045, 0.0066], + [-0.0209, -0.0081, 0.0086, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0054, 0.0354, -0.0082, ..., 0.0117, 0.0356, -0.0186], + [ 0.0513, 0.0141, -0.0010, ..., 0.0376, 0.0532, 0.0026], + [ 0.0175, -0.0347, -0.0013, ..., -0.0063, 0.0057, -0.0283], + ..., + [ 0.0354, -0.0048, -0.0209, ..., -0.0153, 0.0262, -0.0113], + [ 0.0037, 0.0244, 0.0356, ..., -0.0026, 0.0170, 0.0117], + [ 0.0486, 0.0175, 0.0043, ..., 0.0016, -0.0033, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0117, ..., -0.0075, -0.0337, 0.0076], + [ 0.0491, 0.0413, 0.0251, ..., -0.0060, 0.0393, 0.0040], + [-0.0162, 0.0039, 0.0135, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0026, 0.0261, 0.0041, ..., -0.0038, 0.0347, 0.0053], + [-0.0254, 0.0068, -0.0012, ..., 0.0096, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0124, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0033, -0.0270, -0.0166, ..., 0.0117, -0.0298, 0.0532], + [-0.0016, 0.0044, -0.0129, ..., -0.0161, 0.0150, 0.0308], + [ 0.0137, 0.0084, 0.0139, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0053, -0.0630, 0.0005, ..., 0.0074, -0.0183, 0.0214], + [ 0.0034, -0.0018, 0.0258, ..., -0.0486, -0.0349, -0.0044], + [ 0.0182, -0.0074, 0.0198, ..., 0.0152, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0036, -0.0028, -0.0006, ..., 0.0035, 0.0022, -0.0117], + [-0.0154, 0.0251, 0.0432, ..., -0.0052, 0.0034, -0.0016], + [-0.0144, -0.0203, -0.0245, ..., -0.0112, -0.0325, -0.0018], + ..., + [ 0.0050, 0.0037, 0.0417, ..., 0.0281, 0.0025, 0.0030], + [-0.0038, -0.0046, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0227, -0.0364, -0.0220, ..., 0.0062, -0.0039, -0.0091]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0156, -0.0078, -0.0481, ..., 0.0449, 0.0212, 0.0127], + [ 0.0154, -0.0227, 0.0220, ..., -0.0221, -0.0020, 0.0086], + [-0.0151, -0.0095, 0.0430, ..., -0.0101, 0.0327, -0.0112], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0146], + [-0.0025, 0.0166, -0.0063, ..., 0.0002, 0.0081, -0.0078], + [-0.0112, -0.0015, 0.0123, ..., 0.0042, -0.0036, -0.0082]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0025, -0.0386, -0.0123, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0084, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0131, 0.0019, -0.0124, ..., 0.0214, 0.0105, 0.0050], + ..., + [-0.0153, -0.0172, -0.0140, ..., 0.0095, 0.0101, 0.0184], + [-0.0154, -0.0356, 0.0201, ..., -0.0047, -0.0114, 0.0332], + [-0.0396, -0.0119, -0.0092, ..., -0.0284, -0.0142, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0276, 0.0312, 0.0136, ..., -0.0195, 0.0112, -0.0142], + [ 0.0192, 0.0081, -0.0077, ..., 0.0018, -0.0135, -0.0198], + [-0.0248, 0.0100, -0.0139, ..., 0.0090, 0.0172, 0.0061], + ..., + [ 0.0466, -0.0047, 0.0003, ..., -0.0008, 0.0239, 0.0047], + [ 0.0547, -0.0194, -0.0106, ..., -0.0017, 0.0311, -0.0044], + [ 0.0035, 0.0101, -0.0098, ..., 0.0203, 0.0170, 0.0143]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0062, -0.0061, 0.0282, ..., -0.0111, -0.0038, -0.0182], + [ 0.0022, 0.0126, -0.0315, ..., -0.0457, -0.0079, 0.0105], + [-0.0009, 0.0261, 0.0245, ..., 0.0063, 0.0074, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0156, 0.0034, -0.0067], + [ 0.0076, -0.0371, 0.0356, ..., -0.0082, 0.0410, -0.0178], + [ 0.0178, -0.0146, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0078, 0.0208, 0.0276, ..., 0.0136, 0.0151, 0.0083], + [-0.0062, 0.0025, -0.0049, ..., 0.0145, 0.0005, 0.0048], + [ 0.0056, 0.0021, 0.0242, ..., -0.0120, -0.0070, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0059, ..., 0.0248, -0.0095, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., -0.0010, -0.0295, 0.0400], + [-0.0114, 0.0130, 0.0305, ..., -0.0107, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0049, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0146, -0.0043, 0.0288, ..., 0.0150, 0.0583, 0.0001], + [-0.0281, 0.0156, -0.0386, ..., -0.0444, -0.0093, -0.0013], + ..., + [ 0.0030, 0.0157, 0.0217, ..., -0.0005, 0.0115, 0.0151], + [ 0.0081, 0.0195, -0.0283, ..., -0.0009, 0.0021, -0.0342], + [-0.0162, -0.0195, -0.0013, ..., -0.0024, 0.0165, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0146, -0.0138, 0.0459, ..., -0.0018, -0.0320, 0.0278], + [ 0.0019, 0.0045, 0.0270, ..., -0.0220, -0.0183, -0.0271], + [ 0.0221, 0.0217, -0.0201, ..., 0.0030, 0.0062, -0.0053], + ..., + [-0.0258, 0.0034, 0.0215, ..., -0.0018, 0.0237, -0.0266], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0073], + [ 0.0026, -0.0082, -0.0130, ..., -0.0334, 0.0057, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0155, -0.0454, ..., -0.0134, -0.0128, 0.0400], + [ 0.0291, -0.0189, -0.0110, ..., -0.0288, 0.0025, -0.0186], + [ 0.0043, 0.0547, -0.0019, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0275, 0.0042, -0.0269, ..., -0.0096, -0.0287, 0.0150], + [ 0.0101, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0095, 0.0108, ..., -0.0058, 0.0703, -0.0052]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0091, -0.0315, ..., 0.0074, 0.0254, 0.0359], + [ 0.0079, 0.0051, 0.0040, ..., -0.0119, -0.0233, -0.0065], + [ 0.0046, -0.0037, -0.0118, ..., -0.0125, -0.0072, 0.0084], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, -0.0026], + [ 0.0339, -0.0066, 0.0223, ..., 0.0410, 0.0002, 0.0128], + [-0.0049, -0.0439, -0.0486, ..., 0.0137, -0.0129, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0053, 0.0131, -0.0023, ..., -0.0093, -0.0001, -0.0229], + [-0.0287, 0.0229, 0.0075, ..., -0.0137, 0.0045, 0.0222], + [ 0.0065, 0.0288, -0.0189, ..., -0.0173, 0.0311, -0.0262], + ..., + [ 0.0233, 0.0070, 0.0038, ..., -0.0023, 0.0278, 0.0064], + [-0.0109, 0.0005, 0.0151, ..., -0.0175, 0.0444, 0.0425], + [ 0.0062, -0.0068, 0.0093, ..., -0.0040, -0.0344, 0.0226]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[ 0.0059, 0.0125, 0.0190, ..., -0.0095, -0.0198, 0.0142], + [-0.0581, 0.0294, -0.0038, ..., -0.0240, 0.0066, 0.0302], + [ 0.0079, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0017, -0.0289, 0.0537, ..., 0.0065, -0.0056, -0.0093], + [ 0.0053, -0.0080, 0.0209, ..., -0.0035, 0.0090, 0.0186], + [-0.0082, -0.0349, -0.0173, ..., -0.0400, 0.0037, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0178, -0.0259, -0.0102, ..., 0.0090, -0.0137, -0.0405], + [ 0.0063, -0.0095, 0.0266, ..., -0.0400, 0.0047, 0.0391], + [ 0.0059, 0.0359, 0.0283, ..., 0.0295, -0.0233, 0.0212], + ..., + [ 0.0071, 0.0124, 0.0132, ..., -0.0139, -0.0033, 0.0095], + [ 0.0299, -0.0167, -0.0028, ..., -0.0052, -0.0032, -0.0206], + [ 0.0072, -0.0067, -0.0038, ..., -0.0032, -0.0047, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, 0.0056, 0.0116, ..., -0.0032, -0.0211, -0.0064], + [-0.0101, -0.0284, -0.0192, ..., 0.0376, 0.0210, 0.0229], + [ 0.0139, -0.0278, 0.0116, ..., 0.0396, -0.0061, -0.0069], + ..., + [ 0.0029, 0.0034, -0.0311, ..., 0.0344, 0.0062, -0.0125], + [-0.0236, -0.0104, -0.0061, ..., 0.0332, -0.0117, -0.0173], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0138, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[ 0.0014, 0.0166, 0.0006, ..., 0.0273, -0.0352, 0.0152], + [-0.0510, -0.0082, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0074, ..., -0.0388, 0.0004, 0.0398], + ..., + [-0.0160, -0.0195, -0.0021, ..., -0.0015, -0.0051, 0.0033], + [-0.0187, 0.0042, 0.0098, ..., 0.0287, -0.0120, -0.0101], + [-0.0181, 0.0039, 0.0247, ..., -0.0149, 0.0066, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0085, 0.0096, 0.0140, ..., -0.0112, -0.0023, -0.0253], + [-0.0166, 0.0330, -0.0171, ..., 0.0024, -0.0403, -0.0110], + [ 0.0086, 0.0347, 0.0275, ..., 0.0153, 0.0068, 0.0088], + ..., + [ 0.0271, -0.0168, -0.0105, ..., -0.0101, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0058, ..., -0.0040, 0.0006, -0.0243], + [-0.0083, 0.0581, 0.0061, ..., 0.0147, 0.0239, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0124, -0.0410, 0.0128, ..., -0.0254, -0.0033, -0.0173], + [ 0.0476, 0.0315, 0.0141, ..., -0.0148, -0.0069, 0.0049], + [-0.0096, -0.0270, 0.0089, ..., 0.0332, 0.0148, -0.0081], + ..., + [-0.0051, -0.0019, 0.0162, ..., -0.0177, 0.0171, -0.0132], + [ 0.0052, -0.0187, -0.0068, ..., 0.0248, -0.0142, -0.0076], + [-0.0143, -0.0047, 0.0150, ..., -0.0028, -0.0015, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0041, 0.0099, ..., 0.0255, -0.0253, -0.0129], + [-0.0042, -0.0139, -0.0160, ..., -0.0108, -0.0403, -0.0408], + [ 0.0175, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0113, 0.0052, 0.0322, ..., -0.0183, -0.0053, 0.0024], + [ 0.0053, -0.0057, 0.0162, ..., 0.0043, 0.0255, 0.0043], + [ 0.0270, 0.0085, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0075, ..., -0.0532, 0.0104, 0.0029], + [-0.0240, -0.0215, -0.0106, ..., -0.0050, 0.0097, -0.0168], + [ 0.0077, -0.0034, -0.0061, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0071, 0.0172, -0.0106, ..., -0.0300, 0.0005, 0.0435], + [ 0.0052, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0066], + [-0.0167, 0.0125, 0.0102, ..., -0.0247, 0.0120, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0046, -0.0081, 0.0317, ..., 0.0200, -0.0107, -0.0376], + [ 0.0356, 0.0378, 0.0135, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., 0.0027, 0.0050, -0.0322], + ..., + [ 0.0069, 0.0109, -0.0027, ..., 0.0366, -0.0206, 0.0374], + [-0.0054, -0.0080, 0.0038, ..., -0.0096, 0.0086, 0.0192], + [ 0.0099, -0.0178, -0.0038, ..., -0.0166, -0.0042, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0063, -0.0096, -0.0344, ..., -0.0049, 0.0347, -0.0181], + [ 0.0262, -0.0366, -0.0093, ..., 0.0216, 0.0349, -0.0067], + [-0.0100, -0.0286, 0.0056, ..., 0.0061, -0.0282, -0.0006], + ..., + [-0.0128, -0.0016, 0.0043, ..., -0.0089, 0.0165, 0.0076], + [-0.0322, -0.0082, -0.0170, ..., 0.0195, 0.0212, -0.0135], + [-0.0139, 0.0070, -0.0170, ..., -0.0129, 0.0059, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0043, -0.0065, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0115, 0.0374, ..., -0.0153, -0.0188, -0.0138], + ..., + [ 0.0262, 0.0003, -0.0069, ..., 0.0003, -0.0006, -0.0165], + [-0.0222, -0.0156, -0.0437, ..., -0.0065, -0.0165, -0.0282], + [-0.0130, -0.0146, 0.0029, ..., 0.0023, -0.0234, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0144, -0.0649, 0.0102], + [ 0.0249, 0.0205, -0.0050, ..., -0.0064, -0.0091, -0.0103], + [-0.0220, 0.0038, 0.0051, ..., 0.0306, 0.0381, -0.0036], + ..., + [ 0.0200, -0.0042, 0.0170, ..., 0.0118, 0.0320, -0.0157], + [ 0.0349, 0.0121, 0.0197, ..., 0.0398, 0.0315, -0.0217], + [-0.0081, 0.0256, 0.0078, ..., 0.0205, -0.0070, -0.0100]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0065, 0.0199, ..., -0.0063, 0.0276, -0.0312], + [ 0.0178, -0.0055, 0.0096, ..., 0.0113, -0.0398, 0.0293], + [ 0.0430, -0.0120, 0.0085, ..., -0.0029, -0.0067, 0.0060], + ..., + [ 0.0193, -0.0109, -0.0112, ..., -0.0049, 0.0330, 0.0143], + [-0.0122, -0.0542, -0.0205, ..., 0.0034, -0.0159, 0.0320], + [ 0.0303, -0.0039, -0.0425, ..., -0.0064, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0067, -0.0315, ..., 0.0525, -0.0052, -0.0349], + [-0.0031, -0.0032, -0.0156, ..., 0.0055, -0.0240, 0.0305], + [-0.0256, 0.0006, 0.0060, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0053, 0.0231, 0.0209, ..., 0.0073, 0.0050, 0.0461], + [-0.0078, 0.0006, 0.0060, ..., -0.0217, 0.0347, -0.0021], + [ 0.0051, 0.0234, -0.0025, ..., 0.0261, -0.0041, 0.0103]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0037, 0.0006], + [-0.0183, -0.0211, -0.0049, ..., -0.0145, 0.0146, -0.0162], + [-0.0060, -0.0160, 0.0062, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0098, 0.0064, ..., 0.0171, -0.0042, -0.0064], + [-0.0142, 0.0260, -0.0083, ..., 0.0067, 0.0104, -0.0166], + [ 0.0095, -0.0142, -0.0040, ..., 0.0062, 0.0078, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0074, 0.0143, -0.0065, ..., -0.0278, 0.0240, -0.0118], + [ 0.0012, 0.0188, -0.0160, ..., -0.0317, -0.0069, 0.0182], + [ 0.0071, 0.0310, -0.0093, ..., -0.0226, 0.0023, 0.0334], + ..., + [-0.0092, 0.0186, -0.0053, ..., 0.0228, 0.0063, 0.0050], + [ 0.0271, 0.0123, 0.0354, ..., -0.0062, 0.0198, 0.0143], + [ 0.0282, 0.0171, 0.0317, ..., -0.0113, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0110, -0.0265, -0.0115, ..., 0.0049, 0.0038, 0.0132], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0085], + [-0.0044, -0.0190, -0.0063, ..., 0.0166, -0.0135, 0.0137], + ..., + [-0.0200, -0.0009, 0.0057, ..., -0.0339, 0.0361, 0.0006], + [-0.0082, -0.0250, 0.0151, ..., -0.0035, -0.0162, 0.0220], + [-0.0075, 0.0022, 0.0205, ..., -0.0065, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0066, 0.0240, -0.0217, ..., 0.0172, 0.0045, 0.0109], + [-0.0386, 0.0103, 0.0045, ..., -0.0055, 0.0322, 0.0087], + [-0.0427, -0.0052, -0.0042, ..., 0.0062, 0.0110, -0.0135], + ..., + [-0.0085, -0.0044, -0.0264, ..., -0.0586, -0.0630, -0.0129], + [ 0.0042, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0106, -0.0610, 0.0162, ..., -0.0291, -0.0166, -0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0122, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0114], + [ 0.0112, -0.0101, -0.0327, ..., 0.0447, 0.0074, 0.0076], + [ 0.0053, -0.0175, 0.0070, ..., -0.0082, 0.0004, -0.0060], + ..., + [ 0.0295, -0.0042, -0.0103, ..., 0.0100, -0.0160, -0.0288], + [ 0.0085, 0.0253, -0.0177, ..., 0.0098, 0.0081, 0.0056], + [-0.0139, -0.0064, 0.0136, ..., -0.0128, 0.0277, 0.0086]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0061, -0.0276, -0.0271], + [ 0.0135, 0.0064, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0085, 0.0234, 0.0076, ..., 0.0464, 0.0066, 0.0309], + ..., + [-0.0123, -0.0175, 0.0327, ..., 0.0132, 0.0025, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, -0.0034, -0.0444, ..., 0.0058, -0.0283, 0.0029]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0067, -0.0104, 0.0148, ..., -0.0742, -0.0073, 0.0161], + [-0.0079, -0.0133, -0.0060, ..., 0.0056, 0.0049, 0.0239], + [ 0.0189, 0.0221, -0.0090, ..., 0.0049, 0.0408, 0.0049], + ..., + [ 0.0057, 0.0247, 0.0396, ..., 0.0057, -0.0304, -0.0183], + [ 0.0299, -0.0115, -0.0417, ..., -0.0049, -0.0376, 0.0041], + [-0.0210, 0.0168, -0.0003, ..., 0.0063, 0.0178, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[-0.0018, 0.0242, -0.0083, ..., -0.0245, 0.0064, -0.0058], + [ 0.0190, -0.0127, 0.0332, ..., -0.0146, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0079, ..., 0.0003, -0.0093, 0.0182], + ..., + [-0.0430, -0.0403, -0.0087, ..., 0.0135, -0.0141, -0.0055], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0095, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0061, 0.0122, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0122, ..., -0.0160, -0.0272, 0.0299], + [ 0.0095, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0076, ..., 0.0101, -0.0223, -0.0067], + ..., + [ 0.0192, 0.0292, -0.0026, ..., 0.0176, -0.0197, 0.0059], + [ 0.0265, 0.0063, 0.0067, ..., 0.0286, -0.0018, -0.0267], + [-0.0184, -0.0088, -0.0049, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0083, -0.0159, 0.0084, ..., 0.0198, 0.0059, 0.0277], + [ 0.0172, 0.0039, 0.0039, ..., -0.0035, 0.0021, 0.0022], + [-0.0059, 0.0184, 0.0111, ..., 0.0099, -0.0046, -0.0077], + ..., + [-0.0050, -0.0112, -0.0041, ..., -0.0096, -0.0050, 0.0118], + [-0.0427, -0.0391, -0.0192, ..., 0.0089, -0.0238, -0.0084], + [-0.0042, -0.0106, -0.0222, ..., -0.0164, 0.0342, 0.0090]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0087, -0.0161, 0.0119], + [ 0.0041, 0.0074, 0.0312, ..., 0.0074, 0.0286, 0.0317], + [ 0.0214, -0.0145, 0.0229, ..., -0.0082, 0.0056, 0.0277], + ..., + [ 0.0170, -0.0082, 0.0471, ..., 0.0072, 0.0076, -0.0046], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0067, -0.0203], + [ 0.0079, 0.0083, -0.0037, ..., -0.0089, -0.0074, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[ 0.0036, -0.0154, 0.0284, ..., -0.0104, 0.0135, -0.0042], + [ 0.0153, 0.0332, 0.0337, ..., 0.0223, -0.0165, 0.0051], + [-0.0068, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0069, 0.0077, 0.0280, ..., -0.0041, -0.0078, 0.0140], + [-0.0066, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0092, ..., 0.0064, 0.0167, -0.0086]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0079, ..., -0.0238, -0.0099, 0.0118], + [-0.0250, -0.0162, 0.0029, ..., -0.0085, -0.0020, -0.0369], + [ 0.0019, 0.0141, 0.0074, ..., -0.0452, 0.0014, 0.0119], + ..., + [ 0.0149, -0.0140, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0096, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0143, 0.0039, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[-0.0063, -0.0327, -0.0190, ..., 0.0078, 0.0013, -0.0209], + [ 0.0713, 0.0153, -0.0053, ..., -0.0035, 0.0186, 0.0002], + [-0.0227, -0.0094, 0.0188, ..., -0.0128, 0.0193, 0.0068], + ..., + [ 0.0114, 0.0464, 0.0082, ..., 0.0248, 0.0168, -0.0040], + [ 0.0381, -0.0038, 0.0063, ..., -0.0043, -0.0113, -0.0037], + [ 0.0292, 0.0044, 0.0209, ..., -0.0064, 0.0234, 0.0123]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0049, 0.0212, ..., 0.0305, 0.0136, 0.0105], + [-0.0077, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0049], + ..., + [ 0.0084, -0.0275, 0.0225, ..., 0.0055, -0.0189, -0.0105], + [-0.0125, 0.0047, -0.0058, ..., 0.0038, -0.0036, 0.0211], + [-0.0061, 0.0038, -0.0060, ..., -0.0110, 0.0264, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[ 4.6387e-03, -7.4768e-03, -1.0864e-02, ..., 1.8555e-02, + -2.1606e-02, 5.0964e-03], + [-3.8300e-03, 1.2207e-02, -2.3071e-02, ..., 1.2329e-02, + -1.1230e-02, -1.8311e-02], + [ 1.0681e-02, 1.0315e-02, 3.9041e-06, ..., 4.2480e-02, + 3.0884e-02, -2.7344e-02], + ..., + [ 3.3203e-02, 1.2390e-02, 1.4587e-02, ..., 5.3955e-02, + 4.1504e-02, 9.7046e-03], + [ 3.8605e-03, 8.5449e-03, -1.4038e-02, ..., 4.4250e-03, + -1.8921e-02, -2.3682e-02], + [-4.3030e-03, 3.5400e-02, 1.7700e-02, ..., 3.6133e-02, + -3.1128e-02, 6.9885e-03]], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0061, 0.0359, ..., -0.0090, -0.0128, -0.0022], + [ 0.0190, -0.0052, 0.0298, ..., 0.0125, -0.0043, -0.0200], + [ 0.0116, -0.0109, 0.0177, ..., -0.0248, 0.0229, 0.0014], + ..., + [-0.0082, 0.0022, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, -0.0047, -0.0291, ..., 0.0206, 0.0304, -0.0090], + [-0.0072, 0.0245, 0.0047, ..., -0.0413, 0.0216, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0062, -0.0157, -0.0065, ..., -0.0067, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0069], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, -0.0038], + ..., + [ 0.0046, -0.0049, 0.0258, ..., -0.0310, 0.0090, -0.0361], + [ 0.0061, 0.0308, 0.0542, ..., -0.0542, -0.0155, 0.0259], + [-0.0022, 0.0157, -0.0074, ..., -0.0018, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0079, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0107, ..., 0.0103, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0042, ..., 0.0123, -0.0043, 0.0089], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [ 0.0033, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0126, 0.0206, ..., 0.0043, -0.0122, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0096, -0.0231, ..., -0.0214, 0.0447, 0.0136], + [-0.0171, -0.0064, 0.0077, ..., 0.0043, 0.0114, 0.0598], + [-0.0047, 0.0267, -0.0033, ..., 0.0284, -0.0074, -0.0364], + ..., + [-0.0050, -0.0033, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0037, -0.0091, ..., 0.0211, 0.0138, -0.0076], + [ 0.0100, 0.0236, -0.0206, ..., -0.0134, 0.0046, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0064, 0.0294, ..., 0.0033, 0.0270, 0.0055], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0061, -0.0168], + [ 0.0201, -0.0121, 0.0124, ..., -0.0381, -0.0247, -0.0070], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0075], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0066, -0.0466], + [ 0.0327, 0.0284, -0.0048, ..., 0.0145, 0.0140, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0138, -0.0038, -0.0167, ..., -0.0034, 0.0170, 0.0108], + [ 0.0037, -0.0002, 0.0295, ..., -0.0172, 0.0009, 0.0135], + [ 0.0193, 0.0236, -0.0123, ..., 0.0049, -0.0369, 0.0239], + ..., + [ 0.0153, -0.0413, -0.0125, ..., -0.0354, 0.0059, -0.0320], + [-0.0266, -0.0062, -0.0327, ..., -0.0020, 0.0045, 0.0047], + [ 0.0298, 0.0356, -0.0192, ..., 0.0042, -0.0004, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0077, 0.0059, 0.0317, ..., 0.0049, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0079, -0.0096], + [-0.0166, 0.0073, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0050, 0.0260, ..., 0.0159, 0.0034, -0.0474], + [ 0.0038, 0.0070, -0.0074, ..., 0.0140, 0.0557, 0.0044], + [-0.0203, 0.0273, -0.0075, ..., -0.0330, -0.0016, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0072, ..., 0.0039, -0.0093, -0.0277], + [ 0.0057, -0.0061, -0.0058, ..., -0.0060, 0.0140, -0.0161], + [-0.0047, 0.0168, -0.0055, ..., -0.0092, 0.0124, 0.0036], + ..., + [ 0.0398, 0.0452, 0.0251, ..., -0.0067, -0.0330, -0.0012], + [-0.0620, -0.0040, 0.0371, ..., 0.0410, 0.0041, -0.0173], + [ 0.0126, -0.0039, -0.0270, ..., -0.0043, 0.0388, -0.0103]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0127, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0005, 0.0245], + [-0.0449, 0.0017, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0042, ..., 0.0041, 0.0288, 0.0127], + [ 0.0330, 0.0028, 0.0062, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0078, -0.0033, ..., -0.0060, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0066, 0.0040, 0.0527], + [ 0.0037, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0099, -0.0095, 0.0327], + [ 0.0135, 0.0113, -0.0067, ..., 0.0093, -0.0178, -0.0254], + [-0.0530, -0.0125, 0.0281, ..., -0.0242, -0.0128, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, 0.0029], + [-0.0052, -0.0025, -0.0286, ..., 0.0053, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0052, ..., 0.0018, 0.0083, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0099, 0.0150], + [ 0.0014, 0.0037, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0079, 0.0076, ..., 0.0092, -0.0111, 0.0041]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0075, -0.0038, ..., -0.0220, -0.0145, -0.0047], + [-0.0166, 0.0189, -0.0050, ..., 0.0096, 0.0016, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0043, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0070, ..., -0.0049, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0065, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0125, 0.0289, 0.0125, ..., 0.0135, 0.0065, -0.0105], + [ 0.0188, -0.0197, -0.0054, ..., 0.0284, -0.0342, -0.0066], + [-0.0525, -0.0214, 0.0153, ..., -0.0270, 0.0083, 0.0254], + ..., + [-0.0055, 0.0085, 0.0281, ..., 0.0129, -0.0035, -0.0065], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0046, 0.0033], + [-0.0273, -0.0747, 0.0208, ..., -0.0047, 0.0046, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0056, 0.0148, 0.0334, ..., 0.0205, -0.0273, -0.0020], + [-0.0165, 0.0110, -0.0135, ..., 0.0106, -0.0067, 0.0219], + [ 0.0046, 0.0227, -0.0055, ..., 0.0126, -0.0061, 0.0151], + ..., + [-0.0347, -0.0148, 0.0062, ..., 0.0040, 0.0030, -0.0093], + [-0.0073, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0039, -0.0500, -0.0232, ..., 0.0320, 0.0037, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0052, -0.0364], + [ 0.0220, -0.0026, -0.0115, ..., -0.0043, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0117, ..., 0.0444, 0.0049, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0067, 0.0192, -0.0393], + [-0.0102, -0.0244, 0.0275, ..., 0.0038, 0.0043, 0.0081], + [-0.0170, -0.0427, 0.0132, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0050, -0.0378, -0.0359, ..., 0.0552, -0.0075, -0.0160], + [ 0.0047, 0.0142, 0.0134, ..., 0.0055, -0.0153, 0.0354], + [ 0.0094, 0.0040, 0.0057, ..., 0.0028, -0.0249, -0.0136], + ..., + [ 0.0068, -0.0125, 0.0049, ..., -0.0095, -0.0079, -0.0036], + [-0.0073, 0.0439, 0.0166, ..., 0.0248, 0.0134, -0.0187], + [ 0.0175, 0.0061, 0.0049, ..., 0.0410, -0.0096, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0107, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0142, 0.0038, ..., -0.0035, -0.0244, 0.0356], + [ 0.0071, 0.0165, -0.0200, ..., 0.0033, 0.0237, -0.0078], + ..., + [-0.0197, 0.0054, 0.0386, ..., 0.0181, 0.0109, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, -0.0047, -0.0374], + [-0.0085, -0.0095, 0.0172, ..., 0.0243, -0.0219, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-3.4668e-02, 6.2866e-03, 1.9287e-02, ..., -2.2095e-02, + -1.4771e-02, -2.1118e-02], + [ 1.3062e-02, 1.1658e-02, 2.5269e-02, ..., 2.0386e-02, + -2.3193e-02, -3.2227e-02], + [ 2.5513e-02, -4.8523e-03, 4.1809e-03, ..., -5.7068e-03, + 5.4016e-03, 3.3691e-02], + ..., + [ 4.2236e-02, 2.4414e-03, 3.5156e-02, ..., -3.0396e-02, + -4.7207e-05, -2.0386e-02], + [ 3.0151e-02, 1.5503e-02, 9.9487e-03, ..., -7.2327e-03, + -2.5391e-02, -2.4536e-02], + [-6.4087e-04, -6.2561e-03, -2.0599e-03, ..., 4.1504e-03, + -1.9653e-02, -2.8198e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0042, 0.0266, ..., 0.0155, 0.0187, 0.0047], + [-0.0118, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0036, ..., -0.0172, -0.0092, 0.0231], + ..., + [-0.0295, -0.0136, -0.0036, ..., 0.0042, -0.0048, 0.0101], + [-0.0048, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0078], + [-0.0015, -0.0043, -0.0259, ..., -0.0030, -0.0141, -0.0115]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0040, 0.0064, -0.0140, ..., 0.0242, -0.0040, -0.0386], + [-0.0142, 0.0068, -0.0101, ..., -0.0369, -0.0151, 0.0157], + [-0.0267, -0.0045, -0.0074, ..., -0.0126, 0.0182, -0.0115], + ..., + [ 0.0172, -0.0038, 0.0144, ..., 0.0115, -0.0024, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., -0.0023, -0.0033, -0.0137], + [-0.0219, -0.0396, 0.0039, ..., -0.0054, -0.0037, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-3.3417e-03, 1.6846e-02, 1.9409e-02, ..., 3.4912e-02, + -4.1992e-02, 2.0142e-03], + [ 4.3945e-03, 2.7344e-02, 1.4526e-02, ..., -1.8433e-02, + -3.5248e-03, 1.5442e-02], + [ 1.4465e-02, -3.7231e-03, -7.3242e-03, ..., 2.4536e-02, + 3.5889e-02, 5.8289e-03], + ..., + [ 4.3213e-02, 1.4526e-02, -3.1982e-02, ..., 8.8215e-05, + 4.2114e-03, 2.6611e-02], + [ 2.2949e-02, 1.2573e-02, -1.9897e-02, ..., -7.5989e-03, + -3.1250e-02, -4.5654e-02], + [ 4.9133e-03, 1.0910e-03, -6.8359e-03, ..., 4.9133e-03, + -1.0437e-02, -1.5503e-02]], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0115, 0.0388, 0.0359], + [-0.0042, 0.0121, -0.0019, ..., 0.0123, -0.0046, 0.0068], + [-0.0277, 0.0011, -0.0181, ..., -0.0003, -0.0036, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0035, 0.0513, -0.0109], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0110], + [-0.0282, -0.0046, 0.0119, ..., -0.0055, -0.0038, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0033, -0.0128, ..., 0.0140, 0.0308, 0.0059], + [ 0.0120, -0.0332, -0.0320, ..., -0.0075, -0.0140, -0.0032], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, 0.0029, -0.0311, ..., 0.0124, 0.0059, 0.0081], + [ 0.0040, 0.0198, 0.0141, ..., -0.0222, -0.0143, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, 0.0049, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0090, -0.0035, -0.0027, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0049, ..., 0.0302, -0.0018, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0023, -0.0036]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0028, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0037, -0.0222, 0.0262], + ..., + [-0.0147, 0.0043, 0.0260, ..., -0.0042, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0066, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0035, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0112], + [-0.0278, 0.0041, -0.0374, ..., 0.0276, -0.0003, 0.0151], + [ 0.0089, 0.0237, -0.0039, ..., 0.0527, -0.0031, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0072, -0.0095], + [-0.0067, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0036, -0.0128, ..., 0.0063, -0.0371, -0.0064], + [ 0.0064, -0.0130, -0.0028, ..., -0.0204, -0.0013, 0.0057], + [-0.0112, -0.0195, -0.0053, ..., -0.0542, -0.0094, -0.0059], + ..., + [-0.0205, -0.0082, -0.0037, ..., -0.0199, 0.0044, 0.0058], + [-0.0220, -0.0302, -0.0105, ..., 0.0114, 0.0369, 0.0052], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0134, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[ 0.0017, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0034, 0.0056, 0.0238, ..., -0.0140, -0.0320, 0.0041], + [-0.0193, 0.0364, -0.0035, ..., -0.0076, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0065, -0.0232, -0.0464], + [-0.0107, -0.0298, -0.0195, ..., 0.0109, 0.0312, 0.0151], + [-0.0074, -0.0116, -0.0195, ..., 0.0034, 0.0107, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0149, ..., -0.0371, 0.0400, -0.0165], + [ 0.0047, -0.0239, -0.0123, ..., -0.0062, 0.0243, -0.0396], + [-0.0101, 0.0155, -0.0038, ..., -0.0277, 0.0042, 0.0037], + ..., + [ 0.0112, -0.0047, 0.0515, ..., 0.0327, -0.0120, -0.0077], + [-0.0030, 0.0170, 0.0063, ..., -0.0014, 0.0280, -0.0210], + [-0.0016, 0.0076, 0.0055, ..., 0.0153, -0.0044, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0063, 0.0081, -0.0077, ..., -0.0588, 0.0187, -0.0068], + [-0.0102, 0.0199, -0.0077, ..., 0.0080, -0.0332, -0.0026], + ..., + [-0.0051, -0.0464, 0.0277, ..., -0.0143, 0.0198, 0.0071], + [ 0.0162, -0.0053, 0.0082, ..., -0.0105, -0.0038, 0.0297], + [-0.0269, 0.0129, 0.0164, ..., 0.0086, 0.0176, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0033, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, 0.0028, ..., 0.0142, 0.0128, 0.0439], + [-0.0352, 0.0042, -0.0201, ..., 0.0330, 0.0309, -0.0049], + ..., + [-0.0033, 0.0123, 0.0153, ..., -0.0034, -0.0038, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0039, 0.0175], + [-0.0244, 0.0036, -0.0300, ..., 0.0051, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0046, -0.0019, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0029, ..., 0.0310, 0.0088, -0.0023], + [ 0.0033, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0109, -0.0237, ..., -0.0104, 0.0157, -0.0037], + [-0.0081, 0.0205, 0.0140, ..., 0.0262, -0.0195, 0.0077], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, 0.0040, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0018, 0.0034, -0.0093], + [-0.0009, 0.0019, 0.0398, ..., -0.0304, -0.0029, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0019], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0042, 0.0065, 0.0228, ..., -0.0322, 0.0315, -0.0012], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0050, 0.0291], + [ 0.0135, -0.0014, -0.0339, ..., 0.0074, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, -0.0024], + [-0.0166, -0.0043, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0065, -0.0039, ..., 0.0240, 0.0037, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, -0.0020], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0014, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0011, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0066, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0022], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, -0.0039, 0.0036], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, 0.0027, ..., -0.0164, -0.0020, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0033, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0027, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0056, -0.0231, -0.0255, ..., 0.0038, 0.0295, -0.0057], + [-0.0255, 0.0513, 0.0029, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0031, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0067, -0.0070, -0.0364, ..., -0.0175, 0.0306, -0.0078], + [-0.0051, -0.0195, -0.0012, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0075, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0072, -0.0193, -0.0140], + [-0.0166, 0.0062, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0133, -0.0056, 0.0145, ..., -0.0072, 0.0211, -0.0275], + [ 0.0078, -0.0057, -0.0234, ..., 0.0084, 0.0071, -0.0166], + [ 0.0225, -0.0251, 0.0052, ..., 0.0078, 0.0236, -0.0046], + ..., + [ 0.0186, 0.0143, -0.0046, ..., 0.0272, -0.0083, -0.0096], + [ 0.0077, 0.0035, 0.0072, ..., -0.0260, 0.0267, 0.0151], + [ 0.0103, -0.0076, 0.0212, ..., 0.0195, -0.0157, 0.0121]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0087, -0.0270, 0.0251, ..., 0.0096, 0.0049, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.3.proj.1.weight': tensor([[-0.0172, 0.0071, 0.0757, ..., -0.0610, 0.0327, -0.0320], + [-0.0278, 0.0029, -0.1128, ..., 0.0272, -0.0361, 0.0583], + [-0.0347, 0.0947, -0.0361, ..., -0.0166, 0.0172, -0.0332], + ..., + [ 0.0125, 0.0007, -0.0100, ..., 0.0219, -0.0376, -0.0036], + [-0.0239, -0.0574, 0.0586, ..., 0.0378, 0.0830, 0.0608], + [ 0.0176, -0.0205, 0.0076, ..., 0.0248, 0.0493, 0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_proj.weight': tensor([[-4.1199e-03, 5.4932e-03, 6.3782e-03, ..., 2.3346e-03, + -5.4359e-05, 8.1177e-03], + [-9.6893e-04, 8.4305e-04, -8.3923e-04, ..., -8.9111e-03, + -4.5586e-04, 2.2736e-03], + [ 4.7913e-03, -3.5553e-03, 3.2806e-03, ..., -8.0566e-03, + -4.3335e-03, 1.3809e-03], + ..., + [-2.3193e-03, -4.1504e-03, -7.0496e-03, ..., 1.2268e-02, + 4.0588e-03, -8.0566e-03], + [-9.6436e-03, -7.9346e-03, -1.2634e-02, ..., -1.3123e-02, + -5.9204e-03, -6.7139e-04], + [-1.7929e-03, -8.9111e-03, 6.1646e-03, ..., 9.9487e-03, + -4.2915e-04, 1.8005e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_norm.weight': tensor([ 4.7852e-02, -1.6968e-02, 4.7119e-02, 7.8613e-02, -1.3809e-03, + 4.0513e-08, -1.1362e-16, -1.8396e-21, -3.8598e-17, -1.1689e-19, + -5.3204e-21, 3.8257e-24, -6.4851e-22, -1.0727e-24, -2.7504e-23, + 3.6028e-25, -6.9384e-08, -4.4250e-03, -2.9395e-09, -8.8476e-09, + 7.6834e-09, 1.5924e-19, 2.1973e-02, 3.1281e-04, -5.5908e-02, + 4.2578e-01, -1.0889e-01, -3.9062e-03, 1.7822e-02, 8.7891e-02, + 1.0029e-18, 2.1537e-08, 2.6855e-03, -5.1737e-05, -5.5552e-05, + 5.1880e-04, -4.3074e-09, 8.7486e-14, -4.0118e-23, 2.0195e-26, + -7.4216e-27, 8.4703e-22, 3.2558e-21, -1.4343e-03, 1.9775e-02, + -3.8719e-04, -5.4688e-02, -5.0293e-02, 1.7969e-01, 2.9541e-02, + -5.5542e-03, 8.2397e-04, -3.9062e-03, 3.2596e-07, 7.0690e-17, + -7.6762e-22, 5.8981e-17, 1.3897e-22, -2.2334e-22, 9.1382e-27, + 2.7297e-22, -1.3944e-13, 1.9690e-27, 6.0269e-28, 1.2988e-01, + 1.4746e-01, 3.4570e-01, 7.7820e-03, 1.8311e-03, 2.5728e-08, + 7.2858e-17, 3.7481e-20, -1.4162e-18, -9.3512e-19, -7.4115e-20, + -1.3329e-25, 2.2658e-20, -5.5008e-23, 2.5312e-22, 7.6256e-25, + -1.3318e-07, -7.6771e-05, -2.6375e-06, 3.3424e-11, 2.0210e-07, + -1.2559e-15, -2.5635e-03, -1.6022e-03, 3.4570e-01, -1.5625e-02, + 4.1992e-01, 7.9102e-02, -8.3618e-03, 5.3711e-03, -1.2536e-18, + 1.3132e-07, 6.4087e-04, -9.7156e-06, -3.7193e-05, 5.3406e-05, + -7.4971e-08, 1.2546e-14, -2.7303e-25, -2.2877e-24, -4.2662e-27, + -1.5096e-23, 7.9939e-21, 1.2756e-02, 2.8076e-03, -1.2054e-03, + 3.9844e-01, 2.0703e-01, 1.5430e-01, -1.6113e-01, 7.7515e-03, + -4.9805e-02, -3.1128e-03, 1.5087e-07, 2.9707e-17, -5.7439e-21, + 3.4911e-17, 3.6528e-21, 7.3268e-20, 4.3823e-26, 1.4540e-24, + 1.9895e-11, 7.8065e-24, -1.0057e-25], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.self_attn.k_proj.weight': tensor([[-7.0496e-03, 4.8218e-03, -7.4768e-03, ..., -4.2114e-03, + -3.3264e-03, 2.1515e-03], + [-1.9287e-02, -1.4496e-03, 1.9226e-03, ..., 1.4114e-03, + -5.8289e-03, -1.9531e-03], + [-2.1219e-05, -1.2436e-03, 6.7139e-03, ..., 1.1826e-03, + 1.5488e-03, 8.5449e-03], + ..., + [ 2.5330e-03, 2.9602e-03, -2.6550e-03, ..., -5.2261e-04, + -1.1215e-03, 3.8910e-03], + [ 1.1368e-03, 1.0757e-03, -7.6599e-03, ..., -1.0529e-03, + -1.5354e-04, -1.1063e-03], + [ 1.4648e-03, 1.9684e-03, -3.1738e-03, ..., -5.8594e-03, + -2.7466e-04, 1.3962e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.k_norm.weight': tensor([ 9.4727e-02, 5.6641e-02, -5.5664e-02, 7.6660e-02, 1.5335e-03, + -2.1770e-08, -1.1189e-16, 1.7999e-20, -3.8381e-17, -1.6178e-19, + -8.7350e-21, -3.7740e-24, -2.6073e-21, 1.4605e-24, 1.8301e-23, + 6.4300e-25, -6.6590e-08, 4.1504e-03, 1.0477e-09, 8.8476e-09, + 7.5088e-09, -1.1181e-19, 1.6724e-02, 1.3504e-03, -1.5625e-02, + 4.2383e-01, -1.1816e-01, 7.5684e-02, 3.2349e-03, 4.5538e-05, + -1.4704e-18, 6.2864e-08, -2.6093e-03, -5.1498e-05, -5.5313e-05, + -5.4932e-04, -8.7894e-09, -7.1942e-14, 4.0325e-23, 2.8071e-26, + -6.6643e-27, -8.4372e-22, -5.1351e-21, 1.0193e-02, 1.8311e-02, + 2.2125e-04, 8.5449e-02, 1.2598e-01, 2.0801e-01, 1.6211e-01, + 7.1411e-03, 4.9072e-02, -3.7994e-03, -3.8557e-07, -4.6079e-18, + -1.6874e-21, 5.2313e-18, 4.4668e-22, -9.2644e-22, -1.1763e-26, + 2.6635e-22, -2.5757e-13, -2.0397e-26, -9.4663e-28, 8.1177e-03, + 1.4355e-01, 3.4766e-01, -4.5898e-02, 1.6556e-03, 4.7730e-08, + 5.6379e-17, 3.5999e-20, 1.6263e-18, 9.1480e-19, -7.3692e-20, + -6.8259e-26, 2.0964e-20, 5.5421e-23, -2.6304e-22, 7.8194e-25, + -1.6857e-07, -8.5354e-05, -2.6226e-06, 7.9012e-12, 1.9558e-07, + 1.2976e-15, -1.5564e-03, -3.9291e-04, 3.4570e-01, -1.6479e-02, + 4.1992e-01, 5.7983e-03, 1.7578e-02, 8.7402e-02, -1.7872e-19, + 6.9849e-08, 4.4632e-04, 1.0669e-05, -3.7432e-05, 4.8637e-05, + -7.9628e-08, -1.2379e-14, -1.7933e-25, -2.0550e-24, 4.2914e-27, + 1.5303e-23, -7.9409e-21, 8.2397e-03, 1.1292e-02, 7.4387e-04, + 4.0039e-01, 1.8359e-01, 1.4355e-01, 1.5625e-02, 7.4768e-03, + -1.1597e-03, -3.6621e-03, -8.1491e-08, -3.8164e-17, 5.7175e-21, + -3.7513e-17, -3.4940e-21, -7.1998e-20, 2.3426e-26, -2.2101e-24, + 1.9895e-11, 7.8065e-24, -1.0138e-25], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.self_attn.v_proj.weight': tensor([[-0.0040, 0.0033, -0.0051, ..., -0.0042, 0.0036, -0.0044], + [ 0.0073, -0.0018, 0.0019, ..., -0.0043, -0.0013, -0.0023], + [ 0.0152, 0.0034, -0.0048, ..., 0.0084, -0.0022, -0.0011], + ..., + [ 0.0050, 0.0029, -0.0007, ..., -0.0044, 0.0009, 0.0006], + [ 0.0009, 0.0044, -0.0021, ..., -0.0018, 0.0045, -0.0052], + [ 0.0008, -0.0037, 0.0006, ..., -0.0032, -0.0033, -0.0094]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.output_proj.weight': tensor([[ 0.0129, -0.0291, -0.0254, ..., 0.0038, 0.0062, 0.0021], + [ 0.0141, -0.0053, 0.0117, ..., 0.0043, 0.0077, 0.0028], + [-0.0046, 0.0078, 0.0023, ..., 0.0040, 0.0029, -0.0049], + ..., + [ 0.0094, -0.0002, -0.0016, ..., -0.0038, 0.0046, -0.0112], + [-0.0047, 0.0088, 0.0038, ..., -0.0010, 0.0015, 0.0010], + [-0.0056, 0.0040, -0.0090, ..., 0.0036, 0.0044, -0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_proj.weight': tensor([[-0.0025, -0.0159, 0.0013, ..., 0.0058, -0.0039, 0.0007], + [ 0.0072, 0.0210, 0.0029, ..., -0.0058, 0.0050, -0.0034], + [ 0.0085, 0.0194, 0.0049, ..., -0.0053, 0.0130, -0.0042], + ..., + [ 0.0072, 0.0052, 0.0026, ..., 0.0013, -0.0083, 0.0012], + [-0.0029, -0.0007, -0.0029, ..., -0.0027, 0.0109, -0.0044], + [ 0.0002, -0.0038, 0.0008, ..., -0.0022, 0.0015, 0.0057]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_norm.weight': tensor([ 2.0695e-04, 1.8677e-02, 2.3192e-11, 5.6458e-04, 3.6049e-04, + -9.7156e-06, -6.0499e-06, 2.2125e-03, 3.6377e-02, 2.6611e-02, + 1.4305e-04, 3.8330e-02, 5.5859e-01, 3.2196e-03, 8.4839e-03, + 1.1981e-05, 4.2969e-01, 6.8188e-05, 1.9360e-04, 1.1292e-02, + 3.8528e-04, 1.5640e-03, 9.1309e-02, 1.1902e-02, 4.0283e-03, + 6.3171e-03, 2.3682e-02, 3.0696e-06, 9.1553e-03, 2.3651e-04, + 3.6133e-01, 1.9165e-02, 6.0120e-03, 3.8281e-01, 5.3125e-01, + 9.4986e-04, 8.8882e-04, 2.5879e-02, 1.4258e-01, 8.6060e-03, + 1.6113e-02, 2.0142e-03, -4.4890e-07, 3.3875e-03, 1.8921e-03, + 4.7119e-02, 9.3994e-03, 4.0770e-05, 2.3041e-03, 2.0218e-04, + 1.9932e-04, 1.8552e-06, 1.4156e-06, -1.7762e-05, -3.5400e-03, + 1.3672e-02, 2.6367e-02, 1.3275e-03, 1.4954e-02, 1.4526e-02, + -7.0572e-05, -2.9057e-07, 2.9945e-04, 8.1253e-04, 1.1658e-02, + 4.3640e-03, 1.5640e-03, 2.2070e-01, 1.0803e-02, 2.4780e-02, + 3.9258e-01, 1.5259e-05, 4.6143e-02, 2.0020e-02, 3.7842e-03, + -1.9550e-04, 4.4556e-03, 6.7139e-03, 6.1035e-03, 9.2697e-04, + 9.8267e-03, 1.0071e-02, 1.6016e-01, 1.7929e-03, 2.3127e-05, + 1.1719e-02, -1.5945e-03, 1.1265e-05, 8.9111e-03, 4.7852e-01, + 5.0049e-03, 6.5231e-04, 1.4282e-02, 6.4697e-03, 2.5630e-06, + 4.1406e-01, 1.6928e-05, 1.7929e-04, -1.6964e-13, 6.5613e-03, + 3.2617e-01, -2.6524e-06, -4.1199e-03, 4.0283e-03, 3.4912e-02, + 6.1512e-05, -2.4199e-05, 1.8477e-05, 1.9775e-02, 1.0132e-02, + 1.3504e-03, -3.4332e-05, 9.3994e-03, 4.8218e-03, 2.2430e-03, + 7.1335e-04, 2.0874e-02, 1.7471e-03, -9.5367e-04, 5.3101e-03, + 4.9353e-05, 4.5471e-03, 9.1934e-04, 1.2054e-03, 1.5503e-02, + 1.4210e-04, 2.7161e-03, 1.8501e-04], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_proj.weight': tensor([[-0.0022, 0.0007, -0.0010, ..., -0.0041, 0.0014, -0.0015], + [-0.0013, -0.0021, -0.0042, ..., -0.0038, -0.0054, -0.0001], + [ 0.0043, -0.0015, -0.0052, ..., -0.0038, -0.0040, -0.0036], + ..., + [ 0.0038, -0.0025, 0.0025, ..., 0.0010, 0.0006, -0.0012], + [ 0.0008, 0.0005, -0.0023, ..., -0.0032, 0.0006, 0.0013], + [ 0.0017, 0.0019, -0.0031, ..., -0.0045, -0.0006, 0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_norm.weight': tensor([ 1.6499e-04, 1.8677e-02, -2.0009e-11, 2.2125e-04, 3.4523e-04, + 1.2815e-05, 6.0499e-06, 2.1667e-03, 3.5645e-02, 2.6611e-02, + 1.5354e-04, 3.8574e-02, 5.6250e-01, 3.1738e-03, 8.0566e-03, + 9.1195e-06, 4.2969e-01, 6.8188e-05, -1.1015e-04, 1.1230e-02, + 6.5994e-04, 1.5945e-03, 9.1309e-02, 1.2024e-02, 3.8757e-03, + 6.3171e-03, 2.3438e-02, 2.9057e-06, 8.5449e-03, 2.3651e-04, + 3.6133e-01, 1.8921e-02, 5.9509e-03, 3.8281e-01, 5.3125e-01, + 8.7738e-04, 1.0223e-03, 2.6123e-02, 1.4355e-01, 8.7280e-03, + 1.6113e-02, 2.0142e-03, -4.8429e-07, 3.1891e-03, 1.3733e-03, + 4.6875e-02, 9.3384e-03, -1.4901e-05, 2.3193e-03, -7.9155e-05, + 2.0218e-04, 1.2815e-06, 1.5944e-06, 2.4915e-05, 3.9673e-03, + 1.3550e-02, 2.6245e-02, 1.3504e-03, 1.5259e-02, 1.4587e-02, + -7.7248e-05, -5.5507e-07, -3.1471e-04, 7.4005e-04, 1.1719e-02, + 4.3335e-03, 1.5640e-03, 2.2070e-01, 1.0742e-02, 2.4902e-02, + 3.9258e-01, 1.5259e-05, 4.4434e-02, 2.0020e-02, 3.7994e-03, + 1.9455e-04, 3.6926e-03, 6.6833e-03, 6.1646e-03, 9.1171e-04, + 9.7656e-03, 1.0071e-02, 1.6016e-01, 1.5564e-03, -1.7524e-05, + 1.1719e-02, 1.9379e-03, 1.1861e-05, 8.9722e-03, 4.7852e-01, + 4.9744e-03, 6.1798e-04, 1.4099e-02, 6.4697e-03, 2.2501e-06, + 4.1406e-01, 4.2021e-06, 7.7724e-05, 2.2116e-13, 6.5613e-03, + 3.2812e-01, 2.6673e-06, 4.2114e-03, 4.0283e-03, 3.4912e-02, + 6.2943e-05, 3.9101e-05, 1.8835e-05, 1.9165e-02, 1.0193e-02, + 1.5945e-03, 3.5048e-05, 9.3384e-03, 4.7607e-03, 2.6093e-03, + -6.7139e-04, 2.0874e-02, 1.8539e-03, 1.2131e-03, 5.3101e-03, + 3.6955e-05, 4.6997e-03, 9.1171e-04, 1.1749e-03, 1.5076e-02, + -1.4114e-04, 2.8076e-03, -5.2452e-05], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.v_proj.weight': tensor([[-7.5989e-03, -1.1215e-03, 1.2329e-02, ..., -7.8964e-04, + 5.2490e-03, -1.0742e-02], + [-2.1729e-02, -2.2278e-03, 5.7068e-03, ..., -3.0060e-03, + -8.9111e-03, -2.3956e-03], + [-9.3994e-03, -1.8768e-03, 4.5776e-03, ..., -4.2114e-03, + 1.4771e-02, -1.2085e-02], + ..., + [ 1.0010e-02, -5.7220e-04, 4.4346e-05, ..., -8.2779e-04, + 3.7384e-03, 1.9897e-02], + [-7.6294e-03, 1.6708e-03, 1.0803e-02, ..., 1.3046e-03, + -1.7334e-02, -1.2360e-03], + [-1.6846e-02, -1.8082e-03, 8.1177e-03, ..., -1.2665e-03, + -7.4768e-03, -2.2583e-02]], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.output_proj.weight': tensor([[-0.0013, -0.0039, 0.0039, ..., 0.0077, 0.0042, 0.0098], + [ 0.0042, 0.0090, -0.0073, ..., 0.0137, 0.0020, -0.0074], + [ 0.0077, -0.0078, -0.0010, ..., 0.0038, -0.0059, -0.0006], + ..., + [-0.0054, -0.0056, -0.0043, ..., -0.0075, 0.0052, -0.0055], + [ 0.0081, 0.0005, 0.0091, ..., -0.0078, -0.0079, -0.0040], + [ 0.0175, -0.0094, -0.0039, ..., -0.0023, -0.0044, 0.0064]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_img.weight': tensor([[ 0.0078, 0.0121, 0.0018, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0074, 0.0249, 0.0291], + [-0.0095, 0.0080, -0.0121, ..., -0.0342, -0.0327, 0.0065], + ..., + [ 0.0051, -0.0067, 0.0038, ..., 0.0084, 0.0286, -0.0165], + [ 0.0078, -0.0067, -0.0101, ..., -0.0095, 0.0140, -0.0309], + [ 0.0242, -0.0101, 0.0255, ..., -0.0227, 0.0075, -0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, 0.0018], + [ 0.0041, 0.0120, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0080, ..., -0.0056, -0.0057, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, 0.0045, 0.0052], + [-0.0160, 0.0042, 0.0036, ..., 0.0133, -0.0173, 0.0024], + [ 0.0038, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0054, -0.0364, ..., -0.0225, -0.0145, -0.0052], + [-0.0157, -0.0052, -0.0364, ..., 0.0095, 0.0111, -0.0023], + [ 0.0037, 0.0615, 0.0018, ..., -0.0205, -0.0306, -0.0115], + ..., + [-0.0043, 0.0042, -0.0099, ..., -0.0211, -0.0044, 0.0288], + [ 0.0266, -0.0117, 0.0047, ..., -0.0420, -0.0282, -0.0014], + [ 0.0286, 0.0325, 0.0094, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.mlp.layer1.weight': tensor([[ 0.0036, -0.0023, -0.0013, ..., -0.0089, 0.0038, 0.0076], + [-0.0030, 0.0058, -0.0032, ..., -0.0057, 0.0044, 0.0010], + [-0.0087, 0.0012, -0.0061, ..., -0.0046, 0.0040, 0.0060], + ..., + [-0.0016, -0.0054, 0.0029, ..., -0.0075, -0.0067, 0.0027], + [-0.0028, 0.0018, -0.0085, ..., 0.0128, 0.0232, 0.0038], + [ 0.0034, 0.0015, -0.0031, ..., 0.0009, -0.0068, 0.0004]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.mlp.layer2.weight': tensor([[-2.4109e-03, 7.9346e-03, -9.7656e-03, ..., -1.1215e-03, + 2.0599e-03, 1.2268e-02], + [-1.8005e-03, 6.0425e-03, 3.6011e-03, ..., -1.6846e-02, + -2.3071e-02, 4.7607e-03], + [-7.0035e-06, -8.9722e-03, -4.5166e-03, ..., -5.1575e-03, + -7.6294e-03, -1.2573e-02], + ..., + [ 2.1515e-03, 1.1902e-03, -9.5215e-03, ..., 2.8381e-03, + 8.3008e-03, -7.6904e-03], + [-1.0437e-02, 5.0354e-03, 6.3324e-04, ..., 1.0742e-02, + 3.3447e-02, 5.6152e-03], + [-3.8300e-03, 1.1230e-02, 5.2490e-03, ..., 7.4768e-03, + 5.3406e-03, 2.8610e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0476, -0.0464, -0.0449, ..., 0.0027, 0.0028, 0.0029], + [ 0.0457, 0.0447, 0.0435, ..., -0.0021, -0.0022, -0.0023], + [ 0.0454, 0.0449, 0.0439, ..., -0.0023, -0.0024, -0.0024], + ..., + [ 0.0479, 0.0471, 0.0454, ..., -0.0021, -0.0022, -0.0022], + [ 0.0471, 0.0459, 0.0444, ..., -0.0020, -0.0020, -0.0019], + [ 0.0471, 0.0461, 0.0449, ..., -0.0027, -0.0029, -0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0021, 0.0021, 0.0022, ..., 0.0049, 0.0047, 0.0020], + [-0.0017, 0.0016, 0.0016, ..., 0.0049, 0.0052, 0.0017], + [-0.0059, 0.0059, 0.0054, ..., 0.0074, 0.0082, 0.0057], + ..., + [-0.0452, 0.0525, 0.0515, ..., 0.0515, 0.0488, 0.0452], + [ 0.0047, 0.0081, 0.0080, ..., -0.0068, -0.0065, -0.0046], + [ 0.0081, -0.0093, -0.0092, ..., -0.0067, -0.0059, -0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5847e-03, 5.9509e-03, 7.9346e-03, ..., -6.8665e-04, + -6.1417e-04, -7.8201e-04], + [-3.8086e-02, -3.6621e-02, -3.5889e-02, ..., 3.4332e-04, + 3.5286e-04, 3.6049e-04], + [ 4.2725e-02, 4.1016e-02, 4.0527e-02, ..., -1.5259e-05, + -5.3346e-06, -2.0266e-05], + ..., + [ 3.7842e-02, 3.7109e-02, 3.6377e-02, ..., -4.2152e-04, + -3.5667e-04, -3.9291e-04], + [ 4.9561e-02, 4.8096e-02, 4.7363e-02, ..., -7.9632e-05, + -4.7386e-06, -6.6757e-05], + [ 3.8574e-02, 3.7354e-02, 3.6865e-02, ..., 6.9141e-05, + 1.4782e-04, 9.6321e-05]], dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0070, -0.0059, 0.0048, ..., 0.0062, 0.0064, 0.0053], + [-0.0058, -0.0038, -0.0053, ..., 0.0038, 0.0035, 0.0025], + [-0.0043, -0.0044, -0.0085, ..., 0.0043, -0.0074, -0.0071], + ..., + [-0.0078, -0.0134, 0.0259, ..., 0.0132, 0.0273, 0.0216], + [ 0.0114, 0.0058, 0.0103, ..., -0.0056, 0.0052, 0.0107], + [-0.0127, 0.0037, -0.0052, ..., -0.0037, -0.0037, -0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_mlp.1.weight': tensor([[-0.0050, -0.0051, -0.0050, ..., 0.0005, 0.0007, 0.0007], + [ 0.0530, 0.0510, 0.0498, ..., -0.0020, -0.0024, -0.0024], + [ 0.0444, 0.0432, 0.0422, ..., -0.0033, -0.0036, -0.0036], + ..., + [ 0.0894, 0.0869, 0.0859, ..., -0.0005, -0.0007, -0.0007], + [ 0.0050, 0.0049, 0.0050, ..., -0.0005, -0.0006, -0.0005], + [ 0.0640, 0.0625, 0.0605, ..., -0.0010, -0.0010, -0.0010]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_mlp.2.weight': tensor([[-0.0010, 0.0074, 0.0080, ..., 0.0094, 0.0010, 0.0070], + [ 0.0007, 0.0026, 0.0031, ..., 0.0031, -0.0007, 0.0008], + [ 0.0004, 0.0029, 0.0034, ..., 0.0034, -0.0004, 0.0034], + ..., + [ 0.0072, 0.0603, 0.0820, ..., 0.0562, -0.0073, 0.0408], + [ 0.0059, 0.0070, 0.0063, ..., 0.0101, -0.0056, 0.0061], + [-0.0122, -0.0049, -0.0046, ..., -0.0051, 0.0120, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.before_proj.weight': tensor([[ 1.0431e-05, 1.1215e-03, 9.0027e-04, ..., -2.1820e-03, + 2.7657e-04, 3.4027e-03], + [ 3.8147e-03, 5.9509e-04, 1.0681e-02, ..., 1.2207e-02, + -7.5073e-03, -2.5635e-03], + [-1.1215e-03, -3.4637e-03, 2.3346e-03, ..., 1.5030e-03, + -9.9945e-04, -7.0190e-04], + ..., + [ 5.7373e-03, 2.3193e-03, -4.4060e-04, ..., -1.9684e-03, + -1.2207e-04, -5.2490e-03], + [ 2.5635e-03, -4.5166e-03, -4.9133e-03, ..., -1.8692e-03, + 4.3030e-03, -1.2512e-03], + [ 4.6387e-03, 3.4142e-04, -9.4604e-04, ..., 2.0905e-03, + -4.0588e-03, -5.6839e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.0.before_proj.bias': tensor([-0.0016, 0.0080, -0.0006, ..., -0.0053, 0.0027, 0.0032], + dtype=torch.bfloat16), 'net.control_blocks_3.0.after_proj.weight': tensor([[-0.0035, 0.0068, 0.0042, ..., -0.0038, 0.0090, 0.0054], + [-0.0079, -0.0034, 0.0029, ..., -0.0081, 0.0088, -0.0048], + [ 0.0004, -0.0017, -0.0006, ..., -0.0027, 0.0028, -0.0023], + ..., + [ 0.0025, -0.0022, 0.0061, ..., 0.0028, 0.0021, -0.0047], + [ 0.0118, -0.0138, -0.0078, ..., -0.0090, -0.0025, 0.0232], + [ 0.0028, 0.0028, 0.0025, ..., 0.0020, -0.0011, 0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.after_proj.bias': tensor([ 0.0344, -0.0278, 0.0008, ..., -0.0009, -0.0157, 0.0364], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_proj.weight': tensor([[-0.0245, -0.0016, -0.0092, ..., -0.0034, 0.0038, -0.0042], + [-0.0108, -0.0238, -0.0061, ..., -0.0143, -0.0064, -0.0309], + [ 0.0024, 0.0037, -0.0109, ..., 0.0135, 0.0092, -0.0177], + ..., + [ 0.0058, 0.0115, 0.0014, ..., 0.0066, -0.0047, -0.0063], + [ 0.0062, 0.0038, 0.0052, ..., 0.0110, 0.0022, 0.0024], + [-0.0030, -0.0042, 0.0009, ..., -0.0003, -0.0014, 0.0046]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_norm.weight': tensor([ 1.6211e-01, 5.1953e-01, 9.4727e-02, 6.3672e-01, 5.7031e-01, + 4.7265e-08, -2.4872e-03, -2.4011e-10, 2.0339e-13, -1.2306e-28, + 1.0193e-02, 3.3881e-19, -1.1181e-19, -2.1546e-33, 3.7748e-15, + -2.8407e-33, 1.8066e-01, -9.9182e-04, 4.8065e-04, 7.3547e-03, + 1.1921e-05, 5.1117e-04, 1.9727e-01, 7.4219e-01, 7.6172e-01, + 3.8867e-01, -5.8350e-02, 4.3359e-01, -1.3428e-02, 1.6602e-02, + -2.4128e-04, 1.0991e-14, -1.3123e-02, 6.1989e-06, 5.6458e-04, + -3.1433e-03, 3.5390e-08, 2.6090e-15, 4.0531e-05, 4.1567e-13, + 8.4128e-12, -3.3329e-29, 1.7975e-07, 8.5156e-01, 2.4707e-01, + 2.4121e-01, 2.6562e-01, 7.4609e-01, 1.8555e-01, -4.4189e-02, + 1.1902e-02, 1.8359e-01, 1.4305e-04, -1.0315e-02, -1.3924e-04, + 8.0945e-11, 1.6846e-02, -6.3109e-29, 2.7008e-03, -9.2699e-18, + 9.9487e-03, 1.9560e-34, -1.4572e-15, -1.5247e-26, 3.1445e-01, + 1.9922e-01, 4.9414e-01, 1.5527e-01, 9.0332e-02, 2.5495e-08, + 1.1719e-02, 8.4401e-10, 3.4106e-13, -9.5526e-31, -6.5308e-03, + 1.3485e-18, 1.7047e-20, -3.7074e-33, -2.8644e-14, -3.8519e-34, + -5.4932e-03, -2.9206e-05, -1.2131e-03, 2.9755e-03, 6.2943e-05, + 4.2725e-02, 8.1641e-01, 2.4316e-01, 2.2852e-01, 7.5000e-01, + 2.5586e-01, 2.5391e-01, -1.4893e-02, 4.1809e-03, -1.3306e-02, + 2.4203e-14, 1.0193e-02, 2.1338e-05, 3.3140e-05, 3.3417e-03, + 1.4043e-09, -5.7554e-13, 2.3651e-03, -8.5348e-16, 2.1316e-14, + 2.6770e-32, -1.9222e-06, 2.1582e-01, 7.8125e-01, 7.2656e-01, + 7.3438e-01, 2.1973e-01, 3.7305e-01, -5.6076e-04, 2.4414e-01, + 1.1353e-02, 2.4986e-04, -8.4839e-03, -1.3428e-03, 7.3123e-10, + 2.1606e-02, -5.9165e-29, 3.3875e-03, -1.1350e-19, -6.1035e-04, + 1.2488e-34, -9.4369e-15, -1.3545e-23], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.self_attn.k_proj.weight': tensor([[ 0.0018, -0.0083, 0.0069, ..., -0.0095, -0.0078, -0.0113], + [ 0.0317, -0.0110, 0.0075, ..., -0.0096, -0.0066, -0.0208], + [ 0.0042, 0.0084, 0.0037, ..., 0.0049, 0.0028, 0.0038], + ..., + [ 0.0018, -0.0026, -0.0050, ..., -0.0085, -0.0002, 0.0064], + [-0.0094, 0.0029, 0.0021, ..., 0.0157, 0.0093, -0.0103], + [ 0.0045, -0.0045, -0.0019, ..., -0.0126, -0.0001, 0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.k_norm.weight': tensor([ 2.1973e-01, 5.3906e-01, -8.6914e-02, 6.4844e-01, 5.7422e-01, + -3.9581e-08, -8.8501e-03, 4.4929e-10, -1.4477e-13, -5.2854e-29, + -1.1414e-02, -5.6243e-19, -1.0927e-19, -5.6273e-34, 3.0809e-15, + -9.8704e-34, 1.7969e-01, -9.8419e-04, -4.8065e-04, 4.8584e-02, + 1.1861e-05, 7.0095e-05, 2.0508e-01, 7.5000e-01, 7.6562e-01, + 1.4746e-01, 1.9629e-01, 4.9414e-01, -1.1292e-02, 1.6479e-02, + 8.1787e-03, 1.7097e-14, 1.5259e-02, 3.3230e-06, 5.8365e-04, + 3.7842e-02, 3.3993e-08, 5.4401e-15, 4.0293e-05, 4.2100e-13, + -8.4697e-12, -3.3921e-29, -6.0350e-07, 8.4766e-01, 2.3438e-01, + 2.1210e-03, 6.4941e-02, 7.4609e-01, 9.1797e-02, 3.6774e-03, + -2.4609e-01, 7.7637e-02, 1.4019e-04, -9.6436e-03, -1.0300e-04, + 1.2915e-10, 1.7578e-02, -6.1531e-29, -2.4567e-03, -7.9689e-18, + 9.7656e-03, 7.8993e-35, -4.8017e-15, 5.9373e-26, 3.0469e-01, + 2.3535e-01, 4.9805e-01, 1.2988e-01, -9.4238e-02, -3.0734e-08, + -7.0496e-03, 3.2742e-10, 2.4691e-13, 1.4298e-29, 1.6403e-03, + 1.3891e-18, 1.8423e-20, -6.9213e-34, -2.8200e-14, -8.5764e-35, + 4.5776e-03, -4.0293e-05, -1.2054e-03, 2.8076e-03, 6.2466e-05, + 4.2969e-02, 8.1250e-01, 1.2988e-01, 1.7969e-01, 7.6953e-01, + 2.2656e-01, 2.3047e-01, 1.6235e-02, -3.2501e-03, 9.4604e-03, + -2.2760e-14, 5.4626e-03, -2.2054e-05, -4.7386e-06, 3.1494e-02, + 1.3461e-09, -5.2935e-13, -2.4109e-03, -2.1927e-15, 9.3814e-15, + 5.4696e-32, -1.9222e-06, 2.5781e-01, 7.8516e-01, 7.3047e-01, + 7.4219e-01, -5.1270e-02, 4.0625e-01, -4.3701e-02, 5.5542e-03, + -1.5527e-01, 2.5558e-04, 9.0332e-03, -1.3428e-03, -7.2396e-10, + -1.6846e-02, -6.0348e-29, -3.0823e-03, -1.8465e-19, -5.7220e-04, + 9.7049e-35, -8.5487e-15, -1.1529e-23], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.self_attn.v_proj.weight': tensor([[ 5.7068e-03, 4.6387e-03, 6.9275e-03, ..., -1.5198e-02, + 3.6621e-03, 9.0599e-05], + [ 6.2561e-03, -3.5553e-03, 1.4343e-02, ..., -2.3193e-03, + -2.3956e-03, 3.4523e-04], + [ 4.8828e-03, 4.3030e-03, 3.4790e-03, ..., -6.3782e-03, + 5.6763e-03, -7.4463e-03], + ..., + [ 8.7891e-03, -1.0834e-03, 1.0437e-02, ..., 8.3618e-03, + 3.5248e-03, -1.6724e-02], + [-6.5308e-03, 4.5166e-03, 6.0425e-03, ..., 3.4180e-02, + -2.7924e-03, 5.6458e-03], + [-3.3875e-03, 1.0757e-03, -2.8076e-03, ..., 1.3245e-02, + -4.1199e-03, 4.6692e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.output_proj.weight': tensor([[-5.7373e-03, 7.4768e-03, 7.9155e-05, ..., 1.1292e-02, + -1.1719e-02, 7.5073e-03], + [ 1.1047e-02, -3.5553e-03, 1.7578e-02, ..., 4.7913e-03, + -1.2207e-02, 9.5215e-03], + [ 1.9531e-03, 2.9564e-04, 2.3499e-03, ..., 7.7820e-04, + 1.1841e-02, 4.6692e-03], + ..., + [ 1.1047e-02, 4.1504e-03, 1.1749e-03, ..., -4.1809e-03, + -7.9956e-03, -7.2937e-03], + [-7.1106e-03, 8.9722e-03, 4.9438e-03, ..., 4.0588e-03, + -3.7079e-03, 3.6926e-03], + [-1.1475e-02, 9.1553e-03, 1.1597e-02, ..., -2.4719e-03, + 8.1177e-03, -6.1340e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_proj.weight': tensor([[-0.0058, 0.0116, 0.0141, ..., -0.0040, 0.0054, 0.0082], + [ 0.0051, -0.0135, -0.0170, ..., 0.0056, -0.0195, -0.0311], + [ 0.0070, 0.0134, -0.0043, ..., -0.0032, 0.0034, -0.0074], + ..., + [ 0.0020, 0.0212, -0.0030, ..., 0.0063, 0.0024, -0.0049], + [-0.0104, 0.0107, 0.0090, ..., -0.0048, 0.0112, 0.0060], + [ 0.0151, -0.0081, -0.0079, ..., -0.0070, -0.0042, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_norm.weight': tensor([ 1.9775e-02, 4.4727e-01, 2.8320e-02, 4.8578e-06, 3.5048e-05, + 3.2507e-13, 2.3651e-03, 5.1172e-01, 2.6855e-03, 3.7695e-01, + 3.6133e-02, 1.2512e-02, 6.3171e-03, 1.2695e-02, 2.7466e-02, + 1.5381e-02, 1.2207e-02, 7.8201e-05, 2.8163e-06, 3.7109e-02, + 2.5391e-01, 2.5879e-02, 2.4915e-05, 2.4414e-02, 9.4604e-03, + 5.1737e-05, 4.1962e-05, 4.5166e-02, 1.3542e-04, 3.8281e-01, + -4.8399e-05, -1.2684e-04, 5.1880e-03, 8.8215e-05, 2.7008e-03, + 3.2806e-03, 1.9651e-07, 4.1211e-01, 1.0315e-02, 5.4932e-03, + 6.9336e-02, 1.2268e-02, 1.6846e-02, 9.5844e-05, 1.1826e-03, + 1.4526e-02, 3.8910e-03, 2.2217e-02, 2.2461e-02, 1.9775e-02, + 1.1414e-02, 3.8281e-01, 1.2360e-03, 1.9653e-02, 2.6489e-02, + 3.3875e-03, 1.0205e-01, 1.0132e-02, 2.3926e-02, 1.2573e-02, + 1.6975e-04, 5.3101e-03, 1.2878e-02, -1.7738e-04, 5.6152e-03, + 5.5542e-03, -1.0133e-05, 3.3188e-04, 1.6235e-02, 7.9590e-02, + 6.0425e-03, 8.1177e-03, 6.0547e-02, 4.1260e-02, 6.9275e-03, + 5.2185e-03, 7.6771e-05, 2.5879e-02, 3.2227e-02, 2.6489e-02, + 3.9673e-03, 2.0508e-02, 3.1738e-02, -4.5868e-08, 1.9653e-02, + 1.3733e-02, 2.7895e-05, 4.5410e-02, 1.1475e-02, 3.4943e-03, + 1.3916e-02, 7.1335e-04, 5.0049e-03, 2.3193e-02, 1.2878e-02, + -1.3256e-04, 1.3245e-02, 4.0283e-02, 2.0264e-02, 4.8096e-02, + 3.9551e-02, 3.1128e-02, 1.7944e-02, 3.3140e-05, -4.7493e-04, + 3.2616e-04, 1.0864e-02, 2.0264e-02, 1.8799e-02, 1.9287e-02, + 3.7384e-03, 1.3977e-02, 3.6865e-02, 4.9414e-01, -1.6391e-06, + 2.5269e-02, 6.9580e-03, 1.5991e-02, 1.4572e-03, 1.1182e-01, + 5.0068e-06, 2.2411e-05, 3.4571e-05, 1.1475e-02, 1.5015e-02, + 1.5625e-02, 2.3804e-02, 3.1494e-02], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_proj.weight': tensor([[-0.0017, -0.0033, 0.0020, ..., -0.0039, 0.0030, 0.0041], + [ 0.0123, -0.0035, 0.0190, ..., -0.0018, 0.0317, 0.0050], + [ 0.0035, 0.0036, 0.0027, ..., 0.0023, 0.0099, -0.0018], + ..., + [-0.0064, -0.0049, -0.0075, ..., -0.0028, 0.0013, 0.0076], + [-0.0078, -0.0053, -0.0181, ..., -0.0019, 0.0045, 0.0051], + [-0.0024, -0.0004, -0.0070, ..., 0.0022, 0.0089, -0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_norm.weight': tensor([ 2.0142e-02, 4.4727e-01, 2.8198e-02, 4.8876e-06, 3.5048e-05, + -3.1086e-13, 2.3651e-03, 5.1172e-01, 2.7618e-03, 3.7695e-01, + 3.7109e-02, 1.2512e-02, 6.2866e-03, 1.2817e-02, 2.7588e-02, + 1.5259e-02, 1.2451e-02, 7.7248e-05, 3.7402e-06, 3.7354e-02, + 2.5391e-01, 2.5146e-02, 2.4319e-05, 2.4658e-02, 9.5215e-03, + 5.0783e-05, 4.4346e-05, 4.4678e-02, 1.3733e-04, 3.8477e-01, + -4.8637e-05, -1.2684e-04, 5.0659e-03, 8.7738e-05, 2.7313e-03, + 3.3112e-03, 2.0582e-07, 4.1211e-01, 1.0315e-02, 5.4626e-03, + 6.8359e-02, 1.2207e-02, 1.6724e-02, 9.5367e-05, 1.1063e-03, + 1.4526e-02, 3.9368e-03, 2.1851e-02, 2.2461e-02, 2.0020e-02, + 1.1475e-02, 3.8281e-01, 1.1444e-03, 1.9287e-02, 2.6123e-02, + 3.3112e-03, 1.0205e-01, 1.0132e-02, 2.3438e-02, 1.2512e-02, + 1.6785e-04, 5.3406e-03, 1.3000e-02, -1.7738e-04, 5.6152e-03, + 5.4626e-03, -8.5235e-06, 3.1662e-04, 1.6357e-02, 7.9590e-02, + 6.0120e-03, 8.1177e-03, 6.0059e-02, 4.1260e-02, 6.9885e-03, + 5.3406e-03, 7.7248e-05, 2.5513e-02, 3.1738e-02, 2.6489e-02, + 3.9673e-03, 2.0386e-02, 3.1250e-02, -4.4005e-08, 1.9531e-02, + 1.3672e-02, 2.8133e-05, 4.5410e-02, 1.1902e-02, 3.6774e-03, + 1.4282e-02, 7.1716e-04, 5.0049e-03, 2.3071e-02, 1.2756e-02, + -1.3447e-04, 1.2695e-02, 3.9795e-02, 2.0264e-02, 4.8096e-02, + 4.0283e-02, 3.1128e-02, 1.8188e-02, 3.4094e-05, 4.0054e-04, + 3.0518e-04, 1.0925e-02, 1.9775e-02, 1.8921e-02, 1.9165e-02, + 3.7689e-03, 1.3672e-02, 3.6621e-02, 4.9414e-01, -8.8289e-07, + 2.5146e-02, 6.9885e-03, 1.5869e-02, 1.6556e-03, 1.1230e-01, + 2.8312e-06, 2.1338e-05, 3.6478e-05, 1.1719e-02, 1.5076e-02, + 1.5747e-02, 2.3804e-02, 3.1738e-02], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.v_proj.weight': tensor([[-0.0183, 0.0031, 0.0049, ..., 0.0004, -0.0172, -0.0260], + [-0.0047, 0.0093, 0.0052, ..., 0.0046, 0.0053, 0.0051], + [-0.0417, 0.0022, -0.0008, ..., -0.0071, -0.0254, -0.0123], + ..., + [-0.0391, 0.0014, -0.0173, ..., 0.0022, 0.0209, 0.0052], + [ 0.0125, 0.0042, -0.0064, ..., -0.0048, 0.0092, 0.0037], + [ 0.0103, 0.0024, -0.0087, ..., -0.0036, -0.0029, 0.0072]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.output_proj.weight': tensor([[-6.5308e-03, -1.1597e-02, 2.0142e-02, ..., 1.0864e-02, + 1.2268e-02, 5.9509e-03], + [-5.8899e-03, -6.9885e-03, -4.7607e-03, ..., -2.9755e-03, + 5.1270e-03, 1.6724e-02], + [-2.5177e-03, 4.0588e-03, -6.8054e-03, ..., -2.5787e-03, + -2.1210e-03, 3.6926e-03], + ..., + [-3.4790e-03, 1.4954e-03, 6.2561e-03, ..., -6.8054e-03, + 6.3477e-03, 3.5553e-03], + [ 4.3640e-03, 4.1504e-03, 3.9368e-03, ..., -3.3203e-02, + -3.5706e-03, 3.3447e-02], + [ 1.1475e-02, 7.6294e-05, -1.5503e-02, ..., -1.7944e-02, + 3.6011e-03, -4.4556e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_img.weight': tensor([[ 0.0082, -0.0017, 0.0430, ..., -0.0327, -0.0052, 0.0155], + [-0.0278, 0.0280, 0.0249, ..., -0.0072, -0.0378, 0.0271], + [ 0.0134, -0.0121, 0.0312, ..., -0.0077, 0.0253, -0.0113], + ..., + [ 0.0245, -0.0150, 0.0041, ..., -0.0452, 0.0216, -0.0209], + [-0.0081, 0.0069, -0.0471, ..., -0.0106, 0.0157, -0.0082], + [ 0.0062, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0059, -0.0122, -0.0117], + [ 0.0137, 0.0060, -0.0188, ..., 0.0157, 0.0082, -0.0052], + [-0.0306, 0.0012, -0.0272, ..., 0.0100, 0.0337, -0.0050], + ..., + [-0.0040, -0.0046, 0.0044, ..., 0.0209, -0.0107, 0.0071], + [ 0.0193, 0.0210, 0.0123, ..., -0.0146, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0109, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img.weight': tensor([[ 0.0045, -0.0057, 0.0173, ..., 0.0081, 0.0356, 0.0217], + [-0.0498, 0.0062, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0104, 0.0033, -0.0139, ..., -0.0042, 0.0269, -0.0049], + ..., + [ 0.0010, 0.0126, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0073, -0.0162, ..., -0.0057, -0.0073, -0.0120], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0139, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.mlp.layer1.weight': tensor([[-0.0029, 0.0050, -0.0092, ..., 0.0106, 0.0013, -0.0006], + [-0.0026, 0.0053, 0.0047, ..., 0.0059, 0.0116, -0.0123], + [-0.0070, 0.0026, 0.0059, ..., -0.0065, 0.0070, 0.0033], + ..., + [ 0.0194, -0.0036, 0.0075, ..., 0.0078, -0.0006, 0.0038], + [-0.0242, -0.0052, -0.0021, ..., 0.0217, -0.0055, -0.0046], + [ 0.0086, -0.0250, 0.0101, ..., -0.0140, -0.0094, -0.0160]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.mlp.layer2.weight': tensor([[-0.0098, 0.0081, 0.0093, ..., 0.0049, -0.0040, 0.0064], + [-0.0070, 0.0113, 0.0073, ..., -0.0020, 0.0010, -0.0175], + [ 0.0007, 0.0038, -0.0064, ..., 0.0023, 0.0041, -0.0114], + ..., + [-0.0067, -0.0072, 0.0131, ..., -0.0116, -0.0244, 0.0087], + [ 0.0146, 0.0072, -0.0046, ..., -0.0078, 0.0026, -0.0067], + [-0.0046, 0.0062, 0.0068, ..., 0.0039, -0.0120, -0.0143]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0405, -0.0396, -0.0391, ..., 0.0050, 0.0053, 0.0052], + [ 0.0327, 0.0315, 0.0306, ..., -0.0037, -0.0036, -0.0037], + [-0.0417, -0.0398, -0.0396, ..., 0.0050, 0.0051, 0.0051], + ..., + [-0.0447, -0.0435, -0.0422, ..., 0.0034, 0.0036, 0.0037], + [-0.0386, -0.0381, -0.0371, ..., 0.0037, 0.0039, 0.0039], + [-0.0228, -0.0220, -0.0210, ..., 0.0012, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0043, -0.0006, 0.0045, ..., 0.0017, 0.0025, 0.0002], + [-0.0104, 0.0042, -0.0062, ..., -0.0077, -0.0066, -0.0042], + [-0.0049, 0.0018, -0.0061, ..., -0.0084, -0.0084, -0.0020], + ..., + [-0.0220, 0.0037, -0.0267, ..., -0.0286, -0.0270, -0.0038], + [-0.0124, -0.0055, -0.0131, ..., -0.0153, -0.0137, 0.0064], + [ 0.0064, 0.0004, 0.0054, ..., 0.0051, 0.0044, -0.0004]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9570e-02, 5.7861e-02, 5.7129e-02, ..., -1.2159e-04, + -1.6308e-04, -1.1539e-04], + [-4.0283e-03, -3.9978e-03, -3.9978e-03, ..., -4.1008e-05, + -1.1444e-04, -2.2888e-04], + [-1.8555e-02, -1.7944e-02, -1.7090e-02, ..., -6.9809e-04, + -7.0572e-04, -7.1716e-04], + ..., + [-2.9053e-02, -2.8931e-02, -2.8076e-02, ..., -4.1771e-04, + -3.7384e-04, -3.3569e-04], + [-3.3203e-02, -3.2227e-02, -3.0884e-02, ..., -1.8311e-04, + -1.3638e-04, -1.9073e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., -2.0504e-04, + -8.8692e-05, -2.3460e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0089, 0.0040, -0.0204, ..., 0.0052, 0.0135, -0.0127], + [ 0.0073, 0.0063, -0.0064, ..., 0.0061, -0.0039, 0.0053], + [ 0.0121, 0.0049, -0.0063, ..., 0.0080, -0.0047, 0.0102], + ..., + [ 0.0491, 0.0076, -0.0053, ..., 0.0062, -0.0149, 0.0476], + [ 0.0205, -0.0086, 0.0053, ..., -0.0145, -0.0229, 0.0243], + [-0.0029, 0.0096, -0.0164, ..., 0.0114, 0.0080, -0.0063]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_mlp.1.weight': tensor([[-0.0078, -0.0078, -0.0078, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0369, 0.0366, ..., -0.0038, -0.0038, -0.0037], + [ 0.0408, 0.0400, 0.0396, ..., -0.0038, -0.0038, -0.0039], + ..., + [ 0.0262, 0.0253, 0.0254, ..., -0.0021, -0.0021, -0.0020], + [ 0.0057, 0.0063, 0.0064, ..., 0.0014, 0.0014, 0.0013], + [ 0.0166, 0.0170, 0.0167, ..., -0.0025, -0.0025, -0.0026]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_mlp.2.weight': tensor([[-3.2654e-03, 6.0120e-03, 4.9133e-03, ..., 4.3640e-03, + 3.2959e-03, 3.2959e-03], + [ 7.0190e-03, 4.7607e-03, 2.6093e-03, ..., -1.2302e-04, + -7.4158e-03, 3.6240e-05], + [ 3.5553e-03, -4.0894e-03, -5.2490e-03, ..., -3.2806e-03, + -3.6621e-03, -3.2501e-03], + ..., + [ 4.0245e-04, 4.3335e-03, -4.7607e-03, ..., -5.8289e-03, + 3.7994e-03, 3.7537e-03], + [ 7.9956e-03, 1.6968e-02, 1.3916e-02, ..., -4.1504e-03, + -7.9956e-03, -3.8147e-03], + [-4.6387e-03, -4.7913e-03, 2.6855e-03, ..., 4.2114e-03, + 4.6387e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.after_proj.weight': tensor([[ 8.1635e-04, 1.4099e-02, 9.2163e-03, ..., -7.5073e-03, + 9.3994e-03, -5.6458e-03], + [-5.2490e-03, -1.0864e-02, -5.1498e-05, ..., 1.3306e-02, + -3.2959e-03, -6.2866e-03], + [ 3.9673e-03, -7.7820e-03, -8.8120e-04, ..., -5.7068e-03, + -1.7944e-02, 6.0120e-03], + ..., + [-5.3101e-03, 6.7749e-03, 2.5757e-02, ..., -2.0020e-02, + -7.4768e-03, -7.4768e-03], + [-1.1902e-02, -5.5237e-03, 1.6632e-03, ..., 1.1230e-02, + 1.0254e-02, -6.3782e-03], + [-4.4556e-03, 7.5378e-03, -4.5776e-03, ..., -6.2561e-03, + -5.9814e-03, 3.9673e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.after_proj.bias': tensor([ 0.0052, 0.0009, 0.0016, ..., 0.0064, 0.0066, -0.0004], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_proj.weight': tensor([[ 0.0437, -0.0240, -0.0189, ..., 0.0144, -0.0232, 0.0200], + [-0.0255, -0.0161, -0.0065, ..., -0.0056, 0.0101, 0.0036], + [ 0.0311, 0.0232, 0.0126, ..., -0.0045, -0.0199, -0.0047], + ..., + [-0.0085, -0.0048, 0.0018, ..., 0.0041, 0.0041, -0.0118], + [ 0.0167, 0.0259, 0.0025, ..., 0.0038, -0.0092, 0.0112], + [-0.0107, -0.0045, 0.0186, ..., -0.0060, 0.0145, -0.0156]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_norm.weight': tensor([ 1.2344e+00, 3.1836e-01, 4.5117e-01, 2.8711e-01, 1.1172e+00, + 7.7515e-03, 1.1841e-02, -4.2455e-13, 2.0874e-02, 4.4861e-03, + -5.4250e-08, 1.9096e-13, 1.0000e+00, -7.9632e-05, -1.8860e-22, + -2.9419e-02, -6.1035e-03, 8.8501e-03, 1.6235e-02, 1.3916e-02, + 1.5820e-01, -1.4782e-04, 5.0000e-01, 1.3359e+00, 6.4062e-01, + 7.3828e-01, 3.2812e-01, 1.6602e-01, 3.5156e-02, -1.2146e-02, + -3.3875e-03, 5.7459e-05, 5.9509e-03, 1.6785e-04, -6.7749e-03, + 3.3936e-02, 1.1816e-01, -9.1553e-05, 2.4780e-02, 1.1504e-05, + 1.4210e-04, -1.5974e-05, 4.5076e-07, 1.0469e+00, 1.2793e-01, + 1.1875e+00, 6.4062e-01, 1.4355e-01, -4.2969e-02, 1.1108e-02, + 1.0620e-02, 1.6846e-02, 1.3867e-01, -8.9645e-05, 1.6499e-04, + 1.0048e-14, 2.2602e-04, -9.8944e-06, -5.5176e-02, -3.8025e-15, + 3.0640e-02, 7.2718e-06, 5.7983e-03, -3.7231e-03, 5.8594e-01, + 1.1562e+00, 1.1328e+00, 1.1406e+00, 3.1250e-01, -1.2634e-02, + 8.2397e-03, 5.3957e-14, 2.1118e-02, -1.2085e-02, -7.3924e-09, + -5.5067e-14, 4.1504e-02, 2.7008e-03, -3.1764e-20, 6.7188e-01, + 9.1406e-01, -9.2773e-03, -5.6744e-05, 2.6611e-02, 1.1621e-01, + 7.7515e-03, 1.2812e+00, 5.3125e-01, 1.1875e+00, 1.2500e+00, + 2.4121e-01, 4.1992e-01, 4.4922e-02, -1.5869e-02, 4.3335e-03, + 5.0068e-05, -2.5269e-02, -3.8147e-03, 1.7578e-02, 3.5645e-02, + 1.2598e-01, -9.8705e-05, 9.7266e-01, -6.0536e-09, 1.2970e-04, + -1.3471e-05, -1.3279e-10, 6.9531e-01, 1.3750e+00, 1.1953e+00, + 1.2969e+00, 3.3789e-01, 2.5879e-02, 2.0142e-02, 2.1973e-02, + 2.7344e-02, 1.3574e-01, -1.0347e-04, 1.6594e-04, 1.6542e-14, + 1.5068e-04, -1.1206e-05, 1.0469e+00, -4.8572e-15, 3.5889e-02, + 1.0058e-06, 5.9375e-01, 1.0547e+00], dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.self_attn.k_proj.weight': tensor([[-0.0256, 0.0096, 0.0081, ..., 0.0226, -0.0024, 0.0109], + [ 0.0042, 0.0034, 0.0022, ..., -0.0066, 0.0105, -0.0008], + [-0.0291, 0.0134, -0.0076, ..., -0.0049, 0.0238, 0.0005], + ..., + [-0.0023, -0.0036, 0.0055, ..., -0.0068, -0.0052, 0.0079], + [ 0.0024, 0.0055, 0.0126, ..., 0.0192, -0.0034, -0.0075], + [ 0.0066, 0.0018, 0.0126, ..., -0.0008, 0.0061, 0.0229]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.k_norm.weight': tensor([ 1.2812e+00, 7.1875e-01, 4.3555e-01, 2.9297e-01, 1.1172e+00, + 1.2939e-02, 1.9653e-02, 2.4158e-13, 2.1484e-02, 7.6904e-03, + -6.1002e-08, 1.9895e-13, 1.0000e+00, -7.9632e-05, -7.0310e-23, + -7.6904e-03, 2.6855e-03, 8.0078e-02, 1.8066e-02, 3.1586e-03, + 5.7678e-03, -1.0490e-04, 5.3125e-01, 1.2188e+00, 2.1191e-01, + 5.8594e-01, 4.0430e-01, 1.5625e-01, 1.0193e-02, -1.8692e-03, + 4.6997e-03, 4.8161e-05, 4.1199e-03, 4.0283e-02, 4.8218e-03, + -4.3640e-03, 4.3335e-03, -3.9816e-05, 2.9907e-02, 1.1981e-05, + 1.6880e-04, -1.2636e-05, -4.5449e-07, 1.1406e+00, 7.5781e-01, + 1.1406e+00, 4.4141e-01, 2.2070e-01, 1.6846e-02, 5.6641e-02, + -9.8267e-03, 2.8992e-03, 1.1414e-02, -2.9683e-05, 6.2943e-05, + -1.3600e-14, 2.4605e-04, -3.6210e-06, -2.2217e-02, -3.5388e-15, + -3.8452e-03, -4.6492e-06, 3.6621e-03, 1.3733e-02, 6.2891e-01, + 1.0859e+00, 1.1484e+00, 1.1172e+00, 2.8516e-01, 1.8311e-03, + -2.2125e-03, -2.4691e-13, 1.6357e-02, 1.8539e-03, -7.6252e-09, + -4.5075e-14, -1.3184e-02, 2.7008e-03, 3.1764e-20, 6.7188e-01, + 9.1406e-01, 6.5430e-02, 1.5926e-04, 3.8147e-03, 8.8501e-03, + 7.9956e-03, 1.2109e+00, 7.9688e-01, 1.3047e+00, 1.2969e+00, + 2.5146e-02, 4.2188e-01, 1.7578e-02, 2.1362e-02, -3.4027e-03, + 3.6001e-05, 2.7954e-02, 4.5410e-02, 2.3193e-02, -8.1177e-03, + -1.8311e-02, -6.0320e-05, 9.7266e-01, 2.7707e-08, 1.6117e-04, + -9.1195e-06, 7.1907e-12, 4.7070e-01, 1.2188e+00, 1.1484e+00, + 1.4141e+00, 3.3789e-01, -3.8330e-02, 1.0205e-01, 2.1729e-02, + 1.2024e-02, 1.1475e-02, -5.6267e-05, 1.5354e-04, -2.2871e-14, + 1.7929e-04, -6.0499e-06, 1.0469e+00, -5.3291e-15, -1.5625e-02, + 1.2964e-06, 5.9375e-01, 1.0547e+00], dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0095, -0.0096, ..., -0.0177, 0.0064, 0.0099], + [-0.0042, -0.0049, -0.0126, ..., -0.0092, 0.0050, 0.0063], + [-0.0137, -0.0091, 0.0016, ..., 0.0320, -0.0079, -0.0100], + ..., + [ 0.0110, 0.0151, -0.0223, ..., -0.0342, -0.0092, -0.0047], + [ 0.0007, 0.0009, 0.0025, ..., 0.0045, 0.0035, -0.0098], + [ 0.0040, -0.0181, -0.0162, ..., 0.0044, -0.0074, -0.0038]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.output_proj.weight': tensor([[ 0.0148, 0.0173, -0.0488, ..., 0.0023, -0.0142, 0.0139], + [-0.0038, -0.0066, -0.0146, ..., 0.0210, 0.0125, -0.0125], + [ 0.0003, 0.0044, 0.0043, ..., 0.0037, 0.0005, 0.0061], + ..., + [ 0.0070, 0.0051, -0.0271, ..., 0.0037, 0.0056, 0.0059], + [ 0.0009, 0.0048, 0.0065, ..., 0.0038, 0.0160, -0.0135], + [ 0.0040, 0.0108, 0.0039, ..., -0.0024, 0.0085, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_proj.weight': tensor([[-0.0053, 0.0062, -0.0029, ..., 0.0040, 0.0022, 0.0024], + [ 0.0075, -0.0033, 0.0025, ..., 0.0078, 0.0012, 0.0060], + [ 0.0046, 0.0026, -0.0014, ..., -0.0047, 0.0080, -0.0036], + ..., + [-0.0021, 0.0029, -0.0014, ..., -0.0053, 0.0069, -0.0017], + [-0.0088, -0.0069, 0.0020, ..., 0.0078, -0.0011, 0.0081], + [-0.0026, -0.0013, -0.0044, ..., -0.0045, 0.0035, -0.0061]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_norm.weight': tensor([ 2.8076e-03, 1.4062e-01, 2.1118e-02, 6.1719e-01, 2.1240e-02, + 1.4526e-02, 1.2695e-02, 5.3406e-03, 2.0386e-02, 1.2305e-01, + 1.7212e-02, 5.6641e-02, 2.6758e-01, 2.2217e-02, 1.7969e-01, + 2.2461e-02, 6.3965e-02, 2.4292e-02, 5.4016e-03, 2.3438e-02, + 1.0681e-02, 1.6895e-01, 2.6562e-01, 2.0410e-01, 2.4902e-02, + 1.5820e-01, 5.5420e-02, 1.1475e-02, 3.8867e-01, 1.2695e-02, + 1.5015e-02, 5.0781e-01, 4.6631e-02, 4.3701e-02, 9.1553e-03, + 2.0020e-02, 8.9355e-02, 5.0391e-01, 5.7422e-01, 1.1841e-02, + 1.1414e-02, 1.0059e-01, 5.2795e-03, 5.5469e-01, 2.9297e-02, + 4.3213e-02, 1.1169e-02, 3.9648e-01, 8.3618e-03, 2.6978e-02, + 6.7871e-02, 1.5747e-02, 2.0264e-02, 5.1953e-01, 2.7313e-03, + 2.8076e-02, 4.7119e-02, 6.7871e-02, 1.1719e-02, 2.0752e-02, + 3.1281e-03, 1.4587e-02, 5.2490e-03, 3.6328e-01, 2.3193e-02, + 1.7700e-03, 2.4902e-02, 6.4453e-02, 5.5664e-02, 9.1553e-03, + 9.4604e-03, 6.2109e-01, -3.6716e-05, 6.0547e-01, 2.0386e-02, + 4.1962e-04, 1.4587e-02, 3.1055e-01, 2.9297e-02, 4.5586e-04, + 8.1055e-02, 8.2397e-03, 5.2246e-02, 9.8877e-03, 1.6235e-02, + 8.0566e-03, 1.1719e-02, 5.1025e-02, 2.0996e-01, 1.1536e-02, + 1.0437e-02, 2.3560e-02, 4.2480e-02, 3.3447e-02, 3.0640e-02, + 1.7456e-02, 1.3550e-02, 2.3560e-02, 5.0537e-02, 2.7466e-02, + 5.0049e-02, 1.9897e-02, 6.2561e-03, 5.2246e-02, 5.9766e-01, + 6.0547e-02, 7.2021e-03, 4.6875e-01, 1.5015e-02, 2.8931e-02, + 1.2085e-02, 1.4587e-02, 7.8201e-05, 7.9102e-02, 4.5117e-01, + 2.0264e-02, 2.4048e-02, 5.6641e-01, 1.0254e-01, 3.1128e-02, + 1.5015e-02, 6.9531e-01, 1.4062e-01, 1.6235e-02, 1.1169e-02, + 7.8125e-03, 2.2461e-02, 1.2634e-02], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_proj.weight': tensor([[ 0.0013, -0.0034, 0.0126, ..., -0.0002, -0.0052, 0.0074], + [-0.0046, 0.0003, -0.0056, ..., 0.0067, -0.0096, -0.0072], + [-0.0070, -0.0016, -0.0056, ..., 0.0168, -0.0151, -0.0118], + ..., + [-0.0018, 0.0046, -0.0001, ..., -0.0026, -0.0044, -0.0068], + [ 0.0043, 0.0029, -0.0074, ..., -0.0056, 0.0054, 0.0016], + [-0.0072, 0.0007, 0.0049, ..., 0.0015, 0.0039, 0.0058]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_norm.weight': tensor([ 2.8381e-03, 1.4062e-01, 2.0996e-02, 6.1719e-01, 2.0996e-02, + 1.4526e-02, 1.2817e-02, 5.2185e-03, 2.0630e-02, 1.2305e-01, + 1.7212e-02, 5.5176e-02, 2.6758e-01, 2.2095e-02, 1.7969e-01, + 2.3193e-02, 6.3477e-02, 2.4414e-02, 5.3406e-03, 2.3315e-02, + 1.0986e-02, 1.6992e-01, 2.6758e-01, 2.0410e-01, 2.4780e-02, + 1.5820e-01, 5.4443e-02, 9.3994e-03, 3.9062e-01, 1.2817e-02, + 1.5198e-02, 5.0781e-01, 4.6631e-02, 4.3701e-02, 9.5215e-03, + 1.9653e-02, 8.8867e-02, 5.0391e-01, 5.7422e-01, 1.1780e-02, + 1.1414e-02, 1.0010e-01, 5.2490e-03, 5.5469e-01, 2.9541e-02, + 4.3213e-02, 1.1475e-02, 3.9648e-01, 8.4839e-03, 2.6855e-02, + 6.7871e-02, 1.6357e-02, 2.0386e-02, 5.1953e-01, 3.1128e-03, + 2.7954e-02, 4.6387e-02, 6.7871e-02, 1.1780e-02, 1.9287e-02, + 3.0060e-03, 1.4526e-02, 5.6152e-03, 3.6328e-01, 2.2949e-02, + 1.9302e-03, 2.4414e-02, 6.4453e-02, 5.5664e-02, 9.0942e-03, + 9.8877e-03, 6.2109e-01, -5.1498e-05, 6.0547e-01, 2.0508e-02, + -1.7548e-04, 1.4648e-02, 3.0859e-01, 3.0029e-02, 6.9046e-04, + 8.1543e-02, 8.1787e-03, 5.1514e-02, 9.7656e-03, 1.5747e-02, + 8.1787e-03, 1.2329e-02, 5.1758e-02, 2.0898e-01, 1.1414e-02, + 1.0498e-02, 2.2583e-02, 4.2725e-02, 3.3203e-02, 3.1494e-02, + 1.7456e-02, 1.3611e-02, 2.3560e-02, 5.0537e-02, 2.7344e-02, + 4.9561e-02, 2.0020e-02, 6.8054e-03, 5.2734e-02, 5.9766e-01, + 6.0547e-02, 7.1411e-03, 4.6875e-01, 1.4587e-02, 2.8809e-02, + 1.2756e-02, 1.4587e-02, 8.0109e-05, 7.9102e-02, 4.5117e-01, + 1.9897e-02, 2.4048e-02, 5.6641e-01, 1.0352e-01, 3.1250e-02, + 1.4771e-02, 6.9922e-01, 1.3965e-01, 1.5259e-02, 1.0803e-02, + 7.6599e-03, 2.2339e-02, 1.2390e-02], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.v_proj.weight': tensor([[ 0.0130, -0.0059, -0.0171, ..., 0.0042, -0.0138, -0.0076], + [-0.0179, -0.0020, -0.0033, ..., -0.0018, -0.0227, 0.0190], + [-0.0127, -0.0049, 0.0033, ..., -0.0045, 0.0310, 0.0281], + ..., + [-0.0030, -0.0030, -0.0121, ..., 0.0048, 0.0171, 0.0043], + [ 0.0278, 0.0032, 0.0175, ..., 0.0035, 0.0116, 0.0164], + [ 0.0128, -0.0001, 0.0049, ..., 0.0053, -0.0192, -0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.output_proj.weight': tensor([[ 0.0125, -0.0001, -0.0154, ..., -0.0190, 0.0078, -0.0012], + [ 0.0089, 0.0038, 0.0125, ..., 0.0066, -0.0090, 0.0024], + [ 0.0059, -0.0033, -0.0007, ..., 0.0007, -0.0078, -0.0057], + ..., + [-0.0114, 0.0038, 0.0079, ..., 0.0052, -0.0232, 0.0073], + [ 0.0097, -0.0165, 0.0049, ..., -0.0265, -0.0133, -0.0056], + [ 0.0042, -0.0167, 0.0177, ..., 0.0317, -0.0025, 0.0086]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0065, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [-0.0032, 0.0150, -0.0159, ..., -0.0039, -0.0054, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0071, -0.0067], + [-0.0337, -0.0040, 0.0054, ..., -0.0041, 0.0172, 0.0231], + [-0.0255, -0.0040, -0.0039, ..., 0.0408, 0.0430, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0017], + [ 0.0198, -0.0208, 0.0128, ..., 0.0090, -0.0337, -0.0050], + [ 0.0302, 0.0101, -0.0029, ..., -0.0304, 0.0199, 0.0045], + ..., + [ 0.0171, -0.0276, -0.0073, ..., 0.0231, 0.0182, -0.0040], + [ 0.0221, 0.0388, 0.0056, ..., 0.0023, -0.0488, -0.0361], + [ 0.0058, 0.0118, -0.0055, ..., 0.0042, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img.weight': tensor([[ 2.7344e-02, -1.9908e-05, 1.1841e-02, ..., 1.4725e-03, + 3.4668e-02, 6.1523e-02], + [-2.5757e-02, 2.6703e-03, -1.9653e-02, ..., 1.3916e-02, + -4.6082e-03, 1.0986e-02], + [-8.2397e-03, -1.6846e-02, -4.2725e-02, ..., -5.8594e-02, + 1.5320e-02, 8.5449e-03], + ..., + [ 9.5825e-03, 1.3184e-02, 1.8433e-02, ..., 1.1047e-02, + 8.9111e-03, 5.1514e-02], + [-3.2806e-03, -2.4658e-02, -2.7832e-02, ..., -8.9722e-03, + -3.8818e-02, 8.3618e-03], + [ 1.9043e-02, -2.6733e-02, -2.1973e-02, ..., -4.5654e-02, + -1.5991e-02, 1.6113e-02]], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.mlp.layer1.weight': tensor([[-0.0068, 0.0051, 0.0058, ..., 0.0002, -0.0018, -0.0094], + [-0.0115, 0.0060, 0.0082, ..., 0.0094, -0.0030, 0.0022], + [ 0.0054, -0.0074, 0.0063, ..., 0.0050, -0.0054, -0.0349], + ..., + [-0.0066, -0.0027, 0.0054, ..., 0.0137, 0.0046, -0.0133], + [ 0.0049, 0.0027, -0.0034, ..., -0.0194, -0.0060, -0.0376], + [ 0.0045, 0.0277, 0.0002, ..., -0.0041, 0.0078, -0.0072]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.mlp.layer2.weight': tensor([[ 0.0056, 0.0057, -0.0063, ..., -0.0060, 0.0172, -0.0059], + [-0.0040, -0.0027, -0.0060, ..., 0.0073, -0.0140, -0.0192], + [ 0.0028, 0.0055, 0.0098, ..., 0.0089, 0.0079, -0.0041], + ..., + [-0.0043, 0.0032, -0.0209, ..., -0.0107, 0.0076, -0.0089], + [-0.0117, 0.0106, 0.0123, ..., -0.0101, 0.0023, -0.0024], + [-0.0011, 0.0063, -0.0120, ..., 0.0049, -0.0094, -0.0374]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0347, 0.0344, 0.0330, ..., -0.0041, -0.0042, -0.0043], + [-0.0334, -0.0322, -0.0317, ..., 0.0043, 0.0047, 0.0046], + [-0.0349, -0.0344, -0.0332, ..., 0.0045, 0.0048, 0.0048], + ..., + [-0.0393, -0.0381, -0.0374, ..., 0.0022, 0.0023, 0.0024], + [-0.0400, -0.0393, -0.0383, ..., 0.0021, 0.0022, 0.0022], + [ 0.0078, 0.0078, 0.0073, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0079, -0.0075, -0.0080, ..., -0.0095, -0.0101, 0.0018], + [ 0.0047, -0.0041, -0.0047, ..., -0.0023, -0.0013, 0.0017], + [ 0.0089, -0.0084, -0.0087, ..., -0.0045, -0.0043, -0.0007], + ..., + [ 0.0383, -0.0422, -0.0413, ..., -0.0222, -0.0182, 0.0082], + [ 0.0093, -0.0092, -0.0087, ..., -0.0067, -0.0067, -0.0113], + [-0.0181, 0.0188, 0.0179, ..., 0.0146, 0.0138, 0.0118]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9326e-02, 5.8350e-02, 5.7373e-02, ..., -1.2589e-04, + -5.5313e-05, -2.0790e-04], + [ 6.0547e-02, 5.8838e-02, 5.7129e-02, ..., 3.7575e-04, + 4.8828e-04, 5.1880e-04], + [-4.9561e-02, -4.7852e-02, -4.6631e-02, ..., -1.2934e-05, + 4.2439e-05, -6.6280e-05], + ..., + [ 3.5156e-02, 3.4180e-02, 3.3447e-02, ..., 1.0538e-04, + 3.7432e-05, 6.6757e-05], + [ 6.2988e-02, 6.2012e-02, 6.0791e-02, ..., 5.6076e-04, + 5.9891e-04, 5.9128e-04], + [-5.1758e-02, -5.1270e-02, -5.0049e-02, ..., -3.5286e-04, + -3.9864e-04, -3.5858e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0064, 0.0097, -0.0022, ..., -0.0041, -0.0103, 0.0048], + [-0.0140, 0.0099, 0.0109, ..., -0.0084, -0.0104, 0.0062], + [ 0.0036, 0.0033, -0.0029, ..., 0.0036, -0.0033, 0.0033], + ..., + [ 0.0182, 0.0115, -0.0118, ..., 0.0065, 0.0120, 0.0056], + [ 0.0114, 0.0291, -0.0128, ..., 0.0134, 0.0383, -0.0232], + [-0.0200, 0.0195, 0.0177, ..., -0.0133, -0.0099, 0.0231]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_mlp.1.weight': tensor([[-0.0757, -0.0742, -0.0718, ..., 0.0029, 0.0032, 0.0031], + [-0.0258, -0.0243, -0.0233, ..., 0.0006, 0.0007, 0.0006], + [ 0.0559, 0.0542, 0.0530, ..., -0.0017, -0.0017, -0.0017], + ..., + [ 0.0430, 0.0420, 0.0415, ..., -0.0019, -0.0018, -0.0016], + [ 0.0347, 0.0330, 0.0317, ..., -0.0009, -0.0008, -0.0009], + [-0.0520, -0.0505, -0.0496, ..., 0.0022, 0.0022, 0.0021]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_mlp.2.weight': tensor([[-0.0266, -0.0076, 0.0162, ..., 0.0093, 0.0091, -0.0166], + [ 0.0042, -0.0098, -0.0094, ..., -0.0092, 0.0098, 0.0092], + [-0.0038, -0.0087, -0.0049, ..., -0.0069, 0.0090, 0.0049], + ..., + [ 0.0608, -0.0090, -0.0262, ..., -0.0195, 0.0083, 0.0270], + [ 0.0187, -0.0172, -0.0186, ..., -0.0171, 0.0171, 0.0179], + [ 0.0054, -0.0179, -0.0103, ..., -0.0129, 0.0184, 0.0098]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.after_proj.weight': tensor([[-0.0054, 0.0061, -0.0078, ..., -0.0073, -0.0055, -0.0062], + [-0.0144, -0.0081, 0.0093, ..., 0.0078, -0.0068, 0.0075], + [ 0.0061, 0.0078, 0.0070, ..., -0.0076, 0.0082, -0.0051], + ..., + [ 0.0074, -0.0090, 0.0143, ..., 0.0154, 0.0192, 0.0157], + [ 0.0014, 0.0089, 0.0025, ..., -0.0200, -0.0118, 0.0017], + [-0.0080, -0.0099, -0.0138, ..., -0.0076, -0.0044, -0.0057]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.after_proj.bias': tensor([ 0.0055, -0.0029, 0.0007, ..., 0.0034, 0.0079, 0.0055], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_proj.weight': tensor([[-0.0192, 0.0198, 0.0232, ..., -0.0031, -0.0097, -0.0026], + [-0.0106, -0.0056, -0.0008, ..., 0.0075, -0.0066, -0.0070], + [-0.0182, -0.0057, 0.0047, ..., -0.0069, -0.0079, 0.0070], + ..., + [-0.0043, 0.0033, -0.0081, ..., -0.0080, 0.0104, 0.0029], + [ 0.0066, -0.0008, -0.0037, ..., -0.0289, -0.0199, -0.0289], + [ 0.0036, -0.0161, 0.0070, ..., 0.0107, -0.0027, -0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_norm.weight': tensor([ 9.3750e-01, 1.4609e+00, 1.3750e+00, 1.3594e+00, 1.3574e-01, + 5.8350e-02, 1.6479e-02, -2.1770e-08, -2.4435e-33, 3.0823e-03, + 5.4016e-03, 1.0840e-01, -6.6223e-03, -1.0315e-02, 1.9989e-03, + 1.4687e-04, 5.9326e-02, 9.1406e-01, 1.0156e+00, 3.9291e-04, + 1.4400e-04, 1.1406e+00, 1.6562e+00, 1.6406e+00, 1.0000e+00, + 8.8672e-01, 1.2969e+00, -1.0376e-02, 4.2439e-05, 6.1989e-05, + -9.3994e-03, 2.6131e-04, 2.0874e-02, 1.0078e+00, 9.3384e-03, + 1.0312e+00, 2.8076e-03, 8.0109e-05, 2.3438e-02, -8.1177e-03, + 5.2527e-07, 1.2207e-04, 1.9908e-05, 1.6797e+00, 1.2266e+00, + 1.2734e+00, 1.3438e+00, 1.4531e+00, 6.7871e-02, 1.7452e-04, + 2.9053e-02, 2.6172e-01, 1.0469e+00, 9.0820e-02, 5.1737e-05, + 1.0596e-01, 2.3804e-03, 1.1621e-01, 2.5630e-05, 6.9580e-03, + 1.9372e-06, -1.8845e-03, 3.7231e-03, 1.0938e+00, 1.5234e+00, + 6.7969e-01, 8.3594e-01, 1.1406e+00, 6.0303e-02, -3.1738e-02, + 3.4637e-03, -4.9127e-08, -3.9482e-33, 2.0752e-02, 1.4709e-02, + 2.5482e-03, 3.1281e-04, -2.6093e-03, -1.6724e-02, 1.3062e-02, + 5.8594e-02, -2.8076e-03, 1.1353e-02, 2.4986e-04, 1.5640e-04, + -6.5002e-03, 6.8359e-01, 6.7578e-01, 1.3281e+00, 1.3438e+00, + 7.0312e-01, 1.0681e-03, -3.3140e-05, 3.1471e-05, 4.6387e-03, + -4.5967e-04, 2.9297e-02, 1.6113e-01, 9.0332e-03, 6.4453e-02, + 1.9932e-04, 3.4809e-05, 1.2207e-01, -7.6771e-05, 5.1498e-05, + -2.6226e-05, 5.1260e-05, 4.9805e-01, 1.4453e+00, 1.3047e+00, + 8.4375e-01, 8.4375e-01, 7.3730e-02, -1.7166e-03, 2.7832e-02, + 1.0547e+00, 5.9570e-02, 9.8047e-01, 4.8161e-05, 2.8839e-03, + 2.6245e-02, 5.2490e-02, 1.2815e-05, 5.5695e-04, 9.3937e-05, + 2.3804e-03, 1.0156e+00, -1.0071e-02], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.self_attn.k_proj.weight': tensor([[-0.0176, -0.0067, -0.0075, ..., -0.0074, 0.0067, -0.0098], + [-0.0050, -0.0165, 0.0178, ..., -0.0085, -0.0135, 0.0146], + [-0.0064, 0.0046, 0.0004, ..., -0.0054, -0.0054, 0.0057], + ..., + [-0.0082, 0.0190, -0.0006, ..., 0.0033, 0.0077, 0.0007], + [ 0.0146, -0.0021, 0.0188, ..., -0.0072, 0.0193, -0.0062], + [-0.0028, -0.0042, 0.0037, ..., -0.0037, 0.0043, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.k_norm.weight': tensor([ 7.0312e-01, 1.4141e+00, 1.3672e+00, 1.3047e+00, 1.3672e-01, + 6.5918e-02, 1.1108e-02, -1.8976e-08, -1.8296e-33, -6.6376e-04, + 5.2185e-03, 1.0938e-01, -1.0681e-02, 3.4668e-02, -1.7929e-03, + 5.9843e-05, 3.9673e-03, 9.1406e-01, 1.0156e+00, 4.6921e-04, + 8.9645e-05, 1.1406e+00, 1.4844e+00, 1.4844e+00, 1.6235e-02, + 2.8516e-01, 1.3281e+00, 1.9302e-03, 4.5538e-05, -5.1498e-05, + -8.7280e-03, 5.8746e-04, -1.9287e-02, 1.0000e+00, 6.2561e-03, + 1.0234e+00, 2.6855e-03, 8.8692e-05, -2.5146e-02, -8.1787e-03, + -1.0505e-06, -1.2589e-04, 1.0550e-05, 1.3203e+00, 5.5859e-01, + 1.2031e+00, 1.3984e+00, 1.4297e+00, -4.9438e-03, -7.7820e-04, + 2.4719e-03, -2.2095e-02, 1.0469e+00, 6.2256e-02, -3.4809e-05, + 1.0596e-01, -1.2970e-03, 1.1621e-01, -7.5400e-06, 7.0801e-02, + -1.7583e-06, 2.1606e-02, -2.1484e-02, 1.0938e+00, 1.5625e+00, + 1.0312e+00, 7.8125e-01, 1.1484e+00, 4.9072e-02, -1.9775e-02, + -1.8799e-02, 5.2620e-08, 4.7426e-33, 2.1240e-02, 2.5391e-02, + 3.3936e-02, 7.2098e-04, 2.7344e-02, -3.0151e-02, -1.3123e-02, + -2.6855e-02, 2.8381e-03, 2.7618e-03, 3.1853e-04, 9.2506e-05, + 5.7678e-03, 1.1875e+00, 1.1484e+00, 1.4297e+00, 1.4141e+00, + 3.0859e-01, 1.1536e-02, -6.0320e-05, -4.3631e-05, 6.0425e-03, + -2.2507e-04, 1.2695e-02, 3.2806e-03, -8.5449e-03, -2.2095e-02, + 9.9182e-05, 5.6267e-05, 1.2158e-01, 3.9339e-05, 6.2943e-05, + 2.7657e-05, 4.7445e-05, 1.2344e+00, 1.6484e+00, 1.2734e+00, + 2.3730e-01, 7.7734e-01, -6.0120e-03, -1.5182e-03, -1.5320e-02, + 1.0547e+00, -2.9053e-02, 9.8438e-01, 4.4823e-05, 7.5989e-03, + 2.6489e-02, 5.2002e-02, 8.0466e-06, 4.1748e-02, 9.4414e-05, + 1.5015e-02, 1.0156e+00, -1.9302e-03], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.self_attn.v_proj.weight': tensor([[ 3.8147e-03, -6.7444e-03, 1.0315e-02, ..., 4.6631e-02, + -9.6893e-04, 2.5330e-03], + [ 1.1963e-02, 2.1973e-02, -8.9722e-03, ..., -2.1240e-02, + -8.8501e-03, -5.7678e-03], + [-2.0294e-03, 1.8677e-02, -4.8218e-03, ..., 1.5625e-02, + -4.4556e-03, 9.4604e-03], + ..., + [ 4.2200e-05, -4.1199e-03, -1.9531e-02, ..., 9.4604e-03, + 1.5381e-02, 3.2349e-03], + [ 2.8229e-03, 1.8433e-02, -5.8899e-03, ..., 3.9978e-03, + 4.2114e-03, -3.4943e-03], + [-9.0332e-03, -8.6060e-03, 5.1880e-03, ..., -6.7139e-03, + -8.2397e-03, -7.4463e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.output_proj.weight': tensor([[ 0.0072, 0.0190, 0.0056, ..., 0.0101, 0.0192, -0.0095], + [ 0.0275, 0.0043, -0.0049, ..., -0.0041, -0.0270, -0.0094], + [ 0.0039, 0.0334, -0.0115, ..., -0.0043, -0.0165, 0.0060], + ..., + [-0.0381, -0.0177, 0.0038, ..., 0.0108, -0.0082, -0.0183], + [ 0.0012, 0.0125, -0.0266, ..., -0.0176, 0.0238, -0.0033], + [-0.0432, 0.0046, 0.0233, ..., 0.0222, -0.0132, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_proj.weight': tensor([[-0.0003, -0.0081, -0.0045, ..., -0.0159, -0.0014, 0.0152], + [ 0.0070, -0.0017, -0.0037, ..., 0.0017, 0.0044, 0.0084], + [ 0.0084, 0.0048, -0.0092, ..., 0.0050, 0.0126, -0.0039], + ..., + [ 0.0060, 0.0171, 0.0045, ..., 0.0056, -0.0106, 0.0049], + [-0.0036, 0.0073, -0.0050, ..., -0.0056, -0.0114, 0.0037], + [ 0.0004, -0.0054, 0.0040, ..., -0.0049, 0.0120, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_norm.weight': tensor([0.1113, 0.0237, 0.4238, 0.0752, 0.1670, 0.0310, 0.1553, 0.1206, 0.5195, + 0.0447, 0.0654, 0.0161, 0.0518, 0.0674, 0.0640, 0.2021, 0.4570, 0.0344, + 0.7539, 0.0364, 0.0327, 0.1689, 0.0193, 0.0894, 0.0623, 0.4355, 0.5234, + 0.2754, 0.0767, 0.0349, 0.0139, 0.2305, 0.6719, 0.2539, 0.0466, 0.0271, + 0.0216, 0.0967, 0.0184, 0.0383, 0.0527, 0.1201, 0.0229, 0.0439, 0.2080, + 0.0498, 0.3555, 0.0260, 0.0554, 0.0211, 0.0258, 0.4023, 0.5312, 0.3789, + 0.1025, 0.5898, 0.5938, 0.0366, 0.1367, 0.0144, 0.0913, 0.4805, 0.0153, + 0.5273, 0.1699, 0.4160, 0.1016, 0.0498, 0.0732, 0.4277, 0.0281, 0.5508, + 0.0251, 0.0442, 0.1533, 0.0908, 0.0243, 0.1113, 0.6406, 0.0398, 0.0952, + 0.0879, 0.0292, 0.0461, 0.4219, 0.0674, 0.0708, 0.0854, 0.0172, 0.1533, + 0.0432, 0.1787, 0.0471, 0.0703, 0.0391, 0.1201, 0.0114, 0.3555, 0.5391, + 0.0503, 0.0549, 0.3242, 0.0339, 0.1309, 0.0398, 0.2285, 0.0156, 0.5742, + 0.5273, 0.0359, 0.0171, 0.7344, 0.0273, 0.0225, 0.4746, 0.2178, 0.3008, + 0.7227, 0.0369, 0.1001, 0.0801, 0.5625, 0.6055, 0.0199, 0.0188, 0.1914, + 0.0374, 0.0408], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_proj.weight': tensor([[-1.9531e-03, 1.8158e-03, 6.0730e-03, ..., -1.7090e-03, + 1.0925e-02, -1.0864e-02], + [ 5.8289e-03, -6.7711e-05, -1.7776e-03, ..., 1.7548e-03, + -3.8452e-03, 1.2390e-02], + [-1.0437e-02, 7.6904e-03, -1.3428e-02, ..., 1.0803e-02, + 5.3406e-03, 4.2114e-03], + ..., + [ 7.5378e-03, 3.1128e-03, -7.1411e-03, ..., -4.4250e-04, + -1.6724e-02, 5.8594e-03], + [-8.6060e-03, -1.8539e-03, -6.6833e-03, ..., 2.8381e-03, + -4.6692e-03, 1.7853e-03], + [ 9.9487e-03, -1.4038e-03, 5.2795e-03, ..., -7.4768e-03, + 7.4463e-03, -5.7983e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_norm.weight': tensor([0.1118, 0.0237, 0.4238, 0.0752, 0.1670, 0.0317, 0.1553, 0.1206, 0.5195, + 0.0452, 0.0664, 0.0161, 0.0518, 0.0674, 0.0640, 0.2041, 0.4590, 0.0344, + 0.7500, 0.0361, 0.0325, 0.1699, 0.0194, 0.0889, 0.0618, 0.4316, 0.5234, + 0.2734, 0.0767, 0.0349, 0.0150, 0.2324, 0.6719, 0.2539, 0.0474, 0.0271, + 0.0214, 0.0967, 0.0186, 0.0378, 0.0525, 0.1211, 0.0233, 0.0444, 0.2080, + 0.0498, 0.3555, 0.0262, 0.0557, 0.0208, 0.0262, 0.4023, 0.5312, 0.3789, + 0.1025, 0.5898, 0.5977, 0.0366, 0.1367, 0.0147, 0.0913, 0.4805, 0.0156, + 0.5273, 0.1689, 0.4160, 0.1011, 0.0498, 0.0732, 0.4297, 0.0282, 0.5508, + 0.0249, 0.0439, 0.1533, 0.0903, 0.0243, 0.1118, 0.6406, 0.0398, 0.0942, + 0.0879, 0.0292, 0.0461, 0.4219, 0.0679, 0.0698, 0.0850, 0.0175, 0.1553, + 0.0442, 0.1777, 0.0471, 0.0708, 0.0393, 0.1206, 0.0114, 0.3555, 0.5391, + 0.0498, 0.0552, 0.3242, 0.0339, 0.1309, 0.0398, 0.2285, 0.0156, 0.5742, + 0.5273, 0.0356, 0.0171, 0.7344, 0.0266, 0.0227, 0.4766, 0.2178, 0.3008, + 0.7188, 0.0364, 0.0981, 0.0801, 0.5625, 0.6055, 0.0198, 0.0188, 0.1914, + 0.0386, 0.0410], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.v_proj.weight': tensor([[ 0.0084, -0.0016, -0.0292, ..., 0.0041, 0.0038, -0.0220], + [ 0.0032, 0.0095, -0.0042, ..., 0.0006, -0.0060, -0.0199], + [ 0.0332, -0.0038, -0.0092, ..., -0.0049, -0.0175, -0.0386], + ..., + [-0.0118, 0.0019, 0.0160, ..., -0.0067, -0.0212, 0.0049], + [-0.0311, -0.0029, -0.0271, ..., -0.0032, 0.0128, -0.0242], + [-0.0053, 0.0071, 0.0076, ..., 0.0010, 0.0092, 0.0043]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0059, -0.0087, ..., -0.0060, 0.0027, 0.0042], + [-0.0173, -0.0145, -0.0193, ..., 0.0061, -0.0162, -0.0192], + [ 0.0070, -0.0029, -0.0039, ..., -0.0020, -0.0012, -0.0052], + ..., + [ 0.0082, -0.0060, 0.0193, ..., 0.0190, -0.0111, -0.0039], + [-0.0043, -0.0080, -0.0049, ..., 0.0244, -0.0148, 0.0232], + [-0.0089, -0.0129, 0.0269, ..., -0.0228, 0.0344, 0.0085]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0049, -0.0066, ..., 0.0266, -0.0278, 0.0076], + ..., + [-0.0217, -0.0410, 0.0488, ..., 0.0013, -0.0027, -0.0255], + [ 0.0015, 0.0066, 0.0024, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0057, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, 0.0027, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0033, 0.0293, ..., 0.0066, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0139, 0.0126, -0.0092], + ..., + [-0.0043, 0.0320, -0.0233, ..., -0.0067, 0.0016, 0.0188], + [ 0.0134, -0.0352, 0.0267, ..., -0.0305, -0.0021, 0.0037], + [-0.0205, 0.0140, -0.0019, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0019, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0057, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0016], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, -0.0013, 0.0015], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0022, ..., -0.0103, 0.0073, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.mlp.layer1.weight': tensor([[ 0.0108, 0.0060, -0.0140, ..., 0.0150, 0.0309, -0.0071], + [-0.0115, 0.0194, 0.0139, ..., -0.0099, -0.0183, 0.0038], + [-0.0079, 0.0284, -0.0168, ..., 0.0104, -0.0039, 0.0002], + ..., + [ 0.0049, 0.0046, 0.0630, ..., 0.0175, 0.0330, 0.0019], + [ 0.0021, -0.0042, -0.0120, ..., -0.0060, -0.0045, -0.0141], + [ 0.0013, -0.0130, -0.0064, ..., -0.0143, -0.0232, -0.0150]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.mlp.layer2.weight': tensor([[-0.0043, -0.0125, -0.0120, ..., -0.0203, 0.0067, -0.0003], + [-0.0059, 0.0171, 0.0068, ..., 0.0078, 0.0133, 0.0211], + [-0.0038, 0.0016, 0.0134, ..., -0.0035, 0.0135, 0.0177], + ..., + [-0.0152, 0.0171, 0.0059, ..., -0.0117, -0.0147, 0.0052], + [ 0.0100, -0.0058, 0.0141, ..., 0.0024, 0.0131, 0.0106], + [-0.0130, -0.0079, 0.0022, ..., 0.0167, 0.0105, 0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0444, -0.0437, -0.0427, ..., 0.0010, 0.0008, 0.0009], + [-0.0413, -0.0403, -0.0393, ..., 0.0027, 0.0028, 0.0027], + [-0.0396, -0.0386, -0.0374, ..., 0.0022, 0.0023, 0.0023], + ..., + [ 0.0498, 0.0481, 0.0469, ..., -0.0036, -0.0037, -0.0037], + [ 0.0398, 0.0386, 0.0371, ..., -0.0031, -0.0031, -0.0030], + [-0.0396, -0.0386, -0.0378, ..., 0.0023, 0.0024, 0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0062, 0.0070, 0.0070, ..., 0.0150, 0.0175, 0.0078], + [-0.0017, -0.0014, -0.0012, ..., -0.0032, -0.0031, -0.0013], + [ 0.0013, -0.0071, -0.0061, ..., 0.0034, 0.0036, -0.0070], + ..., + [-0.0325, -0.0776, -0.0645, ..., 0.0221, 0.0236, -0.0718], + [-0.0105, -0.0099, -0.0103, ..., 0.0039, 0.0040, -0.0096], + [ 0.0082, 0.0050, 0.0067, ..., 0.0162, 0.0187, 0.0062]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_cross_attn.1.weight': tensor([[-8.2520e-02, -8.0566e-02, -7.7637e-02, ..., -1.0586e-04, + -1.2064e-04, -1.1063e-04], + [ 3.3691e-02, 3.2471e-02, 3.1982e-02, ..., 1.9222e-06, + 2.7180e-05, 1.4877e-04], + [ 3.0884e-02, 2.9785e-02, 2.8809e-02, ..., 2.7657e-04, + 3.0136e-04, 2.0695e-04], + ..., + [ 5.0293e-02, 4.8828e-02, 4.8340e-02, ..., -1.7929e-04, + -1.7262e-04, -1.0538e-04], + [ 5.4932e-02, 5.3467e-02, 5.2002e-02, ..., -1.6785e-04, + -1.0729e-04, -1.5450e-04], + [-2.5635e-02, -2.4902e-02, -2.4292e-02, ..., -6.2943e-05, + 9.0122e-05, 1.1253e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0039, -0.0059, 0.0014, ..., 0.0275, 0.0325, 0.0023], + [ 0.0033, -0.0044, 0.0059, ..., -0.0010, 0.0016, 0.0042], + [ 0.0018, -0.0019, 0.0026, ..., -0.0084, -0.0086, -0.0021], + ..., + [-0.0198, 0.0078, 0.0280, ..., 0.0178, 0.0287, -0.0089], + [-0.0222, 0.0126, 0.0347, ..., -0.0042, -0.0026, -0.0033], + [ 0.0179, -0.0112, -0.0062, ..., -0.0096, -0.0111, 0.0176]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_mlp.1.weight': tensor([[-0.0771, -0.0752, -0.0737, ..., 0.0002, 0.0003, 0.0003], + [ 0.0282, 0.0273, 0.0267, ..., 0.0003, 0.0004, 0.0004], + [-0.0991, -0.0977, -0.0962, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0723, -0.0708, -0.0698, ..., 0.0002, 0.0003, 0.0002], + [-0.0540, -0.0527, -0.0518, ..., 0.0003, 0.0001, 0.0001], + [ 0.0703, 0.0693, 0.0679, ..., -0.0002, -0.0002, -0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_mlp.2.weight': tensor([[-0.0166, -0.0019, -0.0118, ..., -0.0186, -0.0221, 0.0200], + [-0.0018, -0.0050, 0.0067, ..., 0.0004, 0.0036, -0.0035], + [-0.0019, 0.0020, -0.0027, ..., -0.0019, -0.0019, 0.0019], + ..., + [-0.0486, -0.0170, -0.0557, ..., -0.0356, -0.0126, 0.0272], + [ 0.0071, -0.0099, 0.0264, ..., 0.0056, 0.0048, -0.0047], + [ 0.0138, 0.0110, 0.0107, ..., 0.0124, 0.0065, -0.0101]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.after_proj.weight': tensor([[ 0.0066, -0.0078, -0.0059, ..., -0.0055, 0.0026, -0.0152], + [ 0.0037, -0.0080, 0.0184, ..., -0.0098, -0.0061, -0.0095], + [ 0.0123, 0.0111, 0.0305, ..., 0.0052, 0.0048, -0.0076], + ..., + [ 0.0073, 0.0186, -0.0070, ..., 0.0038, -0.0143, -0.0076], + [-0.0108, -0.0059, -0.0055, ..., 0.0070, -0.0120, 0.0055], + [-0.0166, 0.0165, -0.0107, ..., -0.0049, 0.0009, 0.0073]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.after_proj.bias': tensor([ 0.0026, -0.0143, -0.0110, ..., 0.0014, -0.0015, 0.0062], + dtype=torch.bfloat16)} +[06-26 07:35:44|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:44|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:35:45|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 3 checkpoint. +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:1015:compile_tokenizer_if_enabled] Tokenizer compilation disabled +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:146:__init__] Found 3 hint keys across all samples +[06-26 07:35:45|WARNING|cosmos_transfer2/inference.py:148:__init__] Loading the multicontrol model. Multicontrol inference is not strictly equal to single control +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/config.yaml +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['sunny_background_background'] +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample sunny_background_background +[06-26 07:35:45|DEBUG|cosmos_transfer2/inference.py:186:_generate_sample] InferenceArguments(name='sunny_background_background' prompt_path=None prompt='The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. Beyond the junction the surroundings open onto a dense modern downtown of tall glass-and-steel skyscrapers and a distant city skyline under a clear sky. The road network with painted lane markings, stop lines and pedestrian crosswalks remains in the foreground, multiple vehicles move through the scene, and the camera follows the moving ego vehicle smoothly. Photorealistic, high detail.' negative_prompt='The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Empty barren horizon, flat low-rise outskirts, blurry melted skyline, distorted warped buildings.' seed=2025 guidance=3 video_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4') max_frames=None context_frame_index=None image_context_path=None num_conditional_frames=1 resolution='720' sigma_max=None num_video_frames_per_chunk=93 num_steps=35 show_control_condition=False show_input=False keep_input_resolution=True edge=EdgeConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4'), control_weight=1.0, mask_path=None, mask_prompt=None, preset_edge_threshold='medium') depth=None vis=BlurConfig(control_path=None, control_weight=0.6, mask_path=None, mask_prompt=None, preset_blur_strength='medium') seg=SegConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/semseg_front.mp4'), control_weight=0.4, mask_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4'), mask_prompt=None, control_prompt=None) guided_generation_mask=None guided_generation_step_threshold=25 guided_generation_foreground_labels=None) +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.json +[06-26 07:35:45|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 07:35:45|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None, 'vis': None, 'vis_mask': None, 'vis_mask_prompt': None, 'seg': '/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/semseg_front.mp4', 'seg_mask': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4', 'seg_mask_prompt': None} +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:472:generate_img2world] Loading input video... +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4 +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:492:generate_img2world] Computing prompt text embeddings... +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:501:generate_img2world] Computing negative prompt text embeddings... +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:513:generate_img2world] Processing image context if available... +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:330:read_and_process_image_context] No image context provided. +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:525:generate_img2world] Loading control inputs... +[06-26 07:35:45|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4 +[06-26 07:35:46|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 07:35:46|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:703:read_and_process_control_input] No vis (blur) control input file found, will compute online.. +[06-26 07:35:46|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/semseg_front.mp4 +[06-26 07:35:46|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 07:35:46|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4 +[06-26 07:35:46|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 07:35:46|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 1/2 +[06-26 07:35:54|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 584845 +[06-26 07:35:54|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 0 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 07:42:34|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 2/2 +[06-26 07:42:42|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 643774 +[06-26 07:42:42|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 1 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 07:49:04|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:736:generate_img2world] Average time per chunk: 395.62849446999826 +[06-26 07:49:04|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 07:49:05|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_edge.mp4 +[06-26 07:49:05|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 07:49:06|INFO|cosmos_transfer2/inference.py:311:_generate_sample] vis control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_vis.mp4 +[06-26 07:49:06|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 07:49:06|INFO|cosmos_transfer2/inference.py:311:_generate_sample] seg control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_seg.mp4 +[06-26 07:49:06|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 07:49:07|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 07:49:08|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.mp4 diff --git a/multicontrol_cookbook_grid/sunny/background/background/spec.json b/multicontrol_cookbook_grid/sunny/background/background/spec.json new file mode 100644 index 0000000000000000000000000000000000000000..2b1b04d1f93ac502059d9d68d91c19f7b9bc6b7e --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/spec.json @@ -0,0 +1,21 @@ +{ + "name": "sunny_background_background", + "prompt": "The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. Beyond the junction the surroundings open onto a dense modern downtown of tall glass-and-steel skyscrapers and a distant city skyline under a clear sky. The road network with painted lane markings, stop lines and pedestrian crosswalks remains in the foreground, multiple vehicles move through the scene, and the camera follows the moving ego vehicle smoothly. Photorealistic, high detail.", + "video_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4", + "guidance": 3, + "num_steps": 35, + "seed": 2025, + "edge": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4", + "control_weight": 1.0 + }, + "seg": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/semseg_front.mp4", + "control_weight": 0.4, + "mask_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4" + }, + "vis": { + "control_weight": 0.6 + }, + "negative_prompt": "The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Empty barren horizon, flat low-rise outskirts, blurry melted skyline, distorted warped buildings." +} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.json b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.json new file mode 100644 index 0000000000000000000000000000000000000000..eccbf16000e98836e501fdd4d710337519fb06bd --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.json @@ -0,0 +1 @@ +{"name":"sunny_background_background","prompt_path":null,"prompt":"The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. Beyond the junction the surroundings open onto a dense modern downtown of tall glass-and-steel skyscrapers and a distant city skyline under a clear sky. The road network with painted lane markings, stop lines and pedestrian crosswalks remains in the foreground, multiple vehicles move through the scene, and the camera follows the moving ego vehicle smoothly. Photorealistic, high detail.","negative_prompt":"The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Empty barren horizon, flat low-rise outskirts, blurry melted skyline, distorted warped buildings.","seed":2025,"guidance":3,"video_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4","max_frames":null,"context_frame_index":null,"image_context_path":null,"num_conditional_frames":1,"resolution":"720","sigma_max":null,"num_video_frames_per_chunk":93,"num_steps":35,"show_control_condition":false,"show_input":false,"keep_input_resolution":true,"edge":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4","control_weight":1.0,"mask_path":null,"mask_prompt":null,"preset_edge_threshold":"medium"},"depth":null,"vis":{"control_path":null,"control_weight":0.6,"mask_path":null,"mask_prompt":null,"preset_blur_strength":"medium"},"seg":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/semseg_front.mp4","control_weight":0.4,"mask_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_masks/sunny/background_mask.mp4","mask_prompt":null,"control_prompt":null},"guided_generation_mask":null,"guided_generation_step_threshold":25,"guided_generation_foreground_labels":null} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.mp4 b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..44f6a5ac96029886c31458d7121ebcacdf4cc33d --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b47d0bf8b779f20461ae30c83e05554a2103eb33f84cdfcc2c1956981c3bfaf2 +size 1765465 diff --git a/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.txt b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.txt new file mode 100644 index 0000000000000000000000000000000000000000..572f9d85061c89e01031ef58092fcf6f3758ae06 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background.txt @@ -0,0 +1 @@ +The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. Beyond the junction the surroundings open onto a dense modern downtown of tall glass-and-steel skyscrapers and a distant city skyline under a clear sky. The road network with painted lane markings, stop lines and pedestrian crosswalks remains in the foreground, multiple vehicles move through the scene, and the camera follows the moving ego vehicle smoothly. Photorealistic, high detail. \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_edge.mp4 b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..517a976ec0a5d8e4a484eb65a5ac5ad8308f733a --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57c1f383a1f61fc52a275e2a3d237bee50a45e7901dc709b401226496891cd0c +size 1406752 diff --git a/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_seg.mp4 b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_seg.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..733a6a809b55c3f4407f38707275c2d71e533957 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_seg.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7cf2e246115d6b29665356090b88064e26d2152af8be7b3fd5a30702e9d637bb +size 841159 diff --git a/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_vis.mp4 b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_vis.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..34f0b6e199ee4029ebdfebc39c17fa0c9132a9f0 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/background/background/sunny_background_background_control_vis.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:282799a9e77adfeeb3c049be5c9a739ce7b15b57f4e9e386fb937053bde1d587 +size 293428 diff --git a/multicontrol_cookbook_grid/sunny/car_color/color/config.yaml b/multicontrol_cookbook_grid/sunny/car_color/color/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..ce13a956a2bf94b1899e4673e96ace3f298f917d --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/car_color/color/config.yaml @@ -0,0 +1,882 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + keys_to_skip_loading: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'False' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataloaders: + image_data: + ratio: '0' + video_data: + dataloader: + batch_size: '1' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '32' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control_and_image_context + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' + use_cache: 'False' + ratio: '1' + video_data_1: + dataloader: + batch_size: '6' + cache_augment_fn: functools.partial(, n=1.8) + cache_size: '32' + concat_size: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control_and_image_context + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + num_video_frames: '1' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' + use_cache: 'False' + ratio: '1' + dataset: + augmentor_name: video_basic_augmentor_v2_with_control + caption_type: t2w_qwen2p5_7b + control_input_type: edge + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_control_conditioner +- ema: power +- tokenizer: wan2pt1_tokenizer +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: vid2vid_2B_control + name: vid2vid_2B_control_720p_t24_control_layer4_cr1pt1_embedding_rectified_flow_with_image_context_with_image_data + project: cosmos_transfer2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + base_load_from: null + conditional_frame_timestep: -1.0 + conditional_frames_probs: + 0: 0.4 + 1: 0.4 + 2: 0.2 + conditioner: + _target_: + control_input_depth: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth + output_key: control_input_depth + control_input_depth_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth_mask + output_key: control_input_depth_mask + control_input_edge: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge + output_key: control_input_edge + control_input_edge_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge_mask + output_key: control_input_edge_mask + control_input_inpaint: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint + output_key: control_input_inpaint + control_input_inpaint_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint_mask + output_key: control_input_inpaint_mask + control_input_seg: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg + output_key: control_input_seg + control_input_seg_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg_mask + output_key: control_input_seg_mask + control_input_vis: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis + output_key: control_input_vis + control_input_vis_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis_mask + output_key: control_input_vis_mask + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + reference_image_context: + _target_: + dropout_rate: '0.5' + input_key: + - images + - video + - image_context + num_token: '256' + output_key: null + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + copy_weight_strategy: first_n + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + hint_keys: edge + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + condition_strategy: spaced + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_image_context_dim: '1152' + extra_per_block_abs_pos_emb: 'False' + img_context_deep_proj: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + mlp_ratio: '4.0' + model_channels: '2048' + num_blocks: '28' + num_control_branches: '1' + num_heads: '16' + num_max_modalities: '8' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + separate_embedders: 'False' + share_q_in_i2v_cross_attn: 'False' + spatial_compression_factor: '8' + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_after_proj_for_multi_branch: 'True' + use_crossattn_projection: 'True' + use_cuda_graphs: 'False' + use_input_hint_block: 'False' + use_wan_fp32_strategy: 'True' + vace_block_every_n: '7' + vace_has_mask: 'False' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: uniform + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: false + use_lora: false + use_reference_image: true + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 8 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: '8.63e-05' + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '100000' + f_max: + - '0.5' + f_min: + - '0.2' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'True' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'False' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + frame_loss_log: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '100' + hit_thres: '300' + save_s3: 'True' + save_s3_every_log_n: '10' + load_base_model: + _target_: + config: null + trainer: null + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + compile_config: + recompile_limit: '8' + use_duck_shape: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '100000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/multicontrol_cookbook_grid/sunny/car_color/color/console.log b/multicontrol_cookbook_grid/sunny/car_color/color/console.log new file mode 100644 index 0000000000000000000000000000000000000000..5c782bec268e18bc5ab3e62f439e2a7e382a63fa --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/car_color/color/console.log @@ -0,0 +1,10 @@ +[06-26 08:02:30|INFO|cosmos_transfer2/inference.py:146:__init__] Found 1 hint keys across all samples +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color/config.yaml +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['sunny_car_color_color'] +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample sunny_car_color_color +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.json +[06-26 08:02:31|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None} +[06-26 08:13:01|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color_control_edge.mp4 +[06-26 08:13:01|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:13:02|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.mp4 diff --git a/multicontrol_cookbook_grid/sunny/car_color/color/debug.log b/multicontrol_cookbook_grid/sunny/car_color/color/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..501b5bb849205c47be2e8b6325d9310dd3854ac9 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/car_color/color/debug.log @@ -0,0 +1,969 @@ +[06-26 08:02:07|DEBUG|packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, training=True, smoke=False, verbose=False, experimental_checkpoints=False) +[06-26 08:02:08|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:02:08|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 08:02:19|CRITICAL|cosmos_transfer2/_src/interactive/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:02:19|CRITICAL|cosmos_transfer2/_src/predict2/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 08:02:20|DEBUG|cosmos_transfer2/inference.py:46:__init__] SetupArguments(output_dir=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color') model='edge' checkpoint_path='s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000' experiment='edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5' config_file='cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/config.py' context_parallel_size=1 disable_guardrails=True offload_guardrail_models=False keep_going=True profile=False benchmark=False compile_tokenizer= enable_parallel_tokenizer=False parallel_tokenizer_grid=(-1, -1))(['edge']) +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1... +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2... +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor... +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1_with_control... +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control... +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor_with_control... +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control_and_image_context... +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_with_control_input... +[06-26 08:02:20|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering hdmap_augmentor_for_local_datasets... +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package ('flash_attn_3_nv'_) was not found. +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version 0.21.0, expected at least 0.21.5.dev9. +[06-26 08:02:20|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_large_scale +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_small_overfit +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.experiment_list +[06-26 08:02:20|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.exp_large_scale +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.experiment_list +[06-26 08:02:20|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2.experiments +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot.cosmos_agibot_posttrain_example +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview.cosmos_multiview_example +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview +[06-26 08:02:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview.cosmos_singleview_example +[06-26 08:02:20|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:02:21|INFO|cosmos_transfer2/_src/imaginaire/utils/misc.py:151:set_random_seed] Using random seed 0. +[06-26 08:02:21|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 08:02:21|WARNING|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[06-26 08:02:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[06-26 08:02:21|INFO|cosmos_transfer2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/f176dc95b4a70f53ce01c4b302851595e7322b00/tokenizer.pth +[06-26 08:02:21|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 0.50 s +[06-26 08:02:25|INFO|cosmos_transfer2/_src/transfer2/networks/siglip2.py:65:get_siglip2_model_processor] +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #4-control_input_edge: + ReMapkey + input key: control_input_edge + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #5-control_input_vis: + ReMapkey + input key: control_input_vis + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #6-control_input_depth: + ReMapkey + input key: control_input_depth + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #7-control_input_seg: + ReMapkey + input key: control_input_seg + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #8-control_input_inpaint: + ReMapkey + input key: control_input_inpaint + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #9-control_input_edge_mask: + ReMapkey + input key: control_input_edge_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge_mask + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #10-control_input_vis_mask: + ReMapkey + input key: control_input_vis_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis_mask + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #11-control_input_depth_mask: + ReMapkey + input key: control_input_depth_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth_mask + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #12-control_input_seg_mask: + ReMapkey + input key: control_input_seg_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg_mask + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #13-control_input_inpaint_mask: + ReMapkey + input key: control_input_inpaint_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint_mask + Dtype: None +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #14-reference_image_context: + SigLip2EmbImgContext + input key: ['images', 'video', 'image_context'] + Param count: 0 + Trainable: None + Dropout rate: 0.5 + Input key: ['images', 'video', 'image_context'] + Output key: ['img_context_emb'] +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:02:25|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:25|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up I2VCrossAttentionFull. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 0.07 s +[06-26 08:02:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 0.84 s +[06-26 08:02:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 4.84 s +[06-26 08:02:26|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 4.84 s +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:569:freeze_base_model] +Freezing base model + +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:607:freeze_base_model] ✓ Unfroze img_context_proj +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 0 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 1 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 2 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 3 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 4 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 5 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 6 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 7 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 8 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 9 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 10 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 11 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 12 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 13 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 14 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 15 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 16 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 17 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 18 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 19 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 20 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 21 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 22 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 23 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 24 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 25 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 26 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:622:freeze_base_model] ✓ Unfroze reference image weights in ControlAwareDiTBlock 27 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 0 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 1 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 2 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:637:freeze_base_model] ✓ Unfroze reference image weights in ControlEncoderDiTBlock 3 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:779:load_base_model] Done loading the base model checkpoint. +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 0-th block weight to control net's 0-th block +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 1-th block weight to control net's 1-th block +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 2-th block weight to control net's 2-th block +[06-26 08:02:26|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 3-th block weight to control net's 3-th block +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[06-26 08:02:26|INFO|cosmos_transfer2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[06-26 08:02:26|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[06-26 08:02:26|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[06-26 08:02:26|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[06-26 08:02:29|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[06-26 08:02:29|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[06-26 08:02:29|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:76:__init__] MinimalV4LVGControlVaceDiT( + (x_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=72, out_features=2048, bias=False) + ) + ) + (pos_embedder): VideoRopePosition3DEmb() + (t_embedder): Sequential( + (0): Timesteps() + (1): TimestepEmbedding( + (linear_1): Linear(in_features=2048, out_features=2048, bias=False) + (activation): SiLU() + (linear_2): Linear(in_features=2048, out_features=6144, bias=False) + ) + ) + (blocks): ModuleList( + (0-27): 28 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlAwareDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + ) + ) + ) + (final_layer): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): FinalLayer( + (layer_norm): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (linear): Linear(in_features=2048, out_features=64, bias=False) + (adaln_modulation): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=4096, bias=False) + ) + ) + ) + ) + ) + (t_embedding_norm): RMSNorm() + (img_context_proj): Sequential( + (0): Linear(in_features=1152, out_features=2048, bias=True) + (1): GELU(approximate='none') + ) + (crossattn_proj): Sequential( + (0): Linear(in_features=100352, out_features=1024, bias=True) + (1): GELU(approximate='none') + ) + (control_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=520, out_features=2048, bias=False) + ) + ) + (control_blocks): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): I2VCrossAttentionFull( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): DotProductAttention( + (flash_attention): FlashAttention() + (fused_attention): FusedAttention() + (unfused_attention): UnfusedDotProductAttention( + (scale_mask_softmax): FusedScaleMaskSoftmax() + (attention_dropout): Dropout(p=0, inplace=False) + ) + ) + (k_img): Linear(in_features=2048, out_features=2048, bias=False) + (v_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img): Linear(in_features=2048, out_features=2048, bias=False) + (q_img_norm): RMSNorm() + (k_img_norm): RMSNorm() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) +) +[06-26 08:02:29|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 8.74 s +[06-26 08:02:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 08:02:29|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 08:02:30|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.0.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.1.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.2.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.attn_op._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.q_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks.3.cross_attn.k_img_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 08:02:30|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=[], unexpected_keys=[], incorrect_shapes=[]) +[06-26 08:02:30|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:271:load_model_state_dict_from_checkpoint] Checkpoint weights loaded from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt: None +[06-26 08:02:30|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:273:load_model_state_dict_from_checkpoint] Checkpoint weights loaded successfully (strict=True). +[06-26 08:02:30|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:1015:compile_tokenizer_if_enabled] Tokenizer compilation disabled +[06-26 08:02:30|INFO|cosmos_transfer2/inference.py:146:__init__] Found 1 hint keys across all samples +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color/config.yaml +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['sunny_car_color_color'] +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample sunny_car_color_color +[06-26 08:02:31|DEBUG|cosmos_transfer2/inference.py:186:_generate_sample] InferenceArguments(name='sunny_car_color_color' prompt_path=None prompt='The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. The vehicles moving through the junction are glossy vivid candy-red, their bright paint catching the sunlight and reflecting the sky. A road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, low-rise buildings and sidewalks line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, crisp reflections, detail.' negative_prompt='The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Dull grey washed-out cars, color bleeding onto road or buildings, melted or deformed vehicle bodies.' seed=2025 guidance=3 video_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4') max_frames=None context_frame_index=None image_context_path=None num_conditional_frames=1 resolution='720' sigma_max=None num_video_frames_per_chunk=93 num_steps=35 show_control_condition=False show_input=False keep_input_resolution=True edge=EdgeConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4'), control_weight=1.0, mask_path=None, mask_prompt=None, preset_edge_threshold='medium') depth=None vis=None seg=None guided_generation_mask=None guided_generation_step_threshold=25 guided_generation_foreground_labels=None) +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.json +[06-26 08:02:31|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 08:02:31|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None} +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:472:generate_img2world] Loading input video... +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4 +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:492:generate_img2world] Computing prompt text embeddings... +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:501:generate_img2world] Computing negative prompt text embeddings... +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:513:generate_img2world] Processing image context if available... +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:330:read_and_process_image_context] No image context provided. +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:525:generate_img2world] Loading control inputs... +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4 +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 1/2 +[06-26 08:02:31|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 584845 +[06-26 08:02:32|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 0 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:02:32|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:255:get_condition_with_negative_prompt] Getting condition with image context +[06-26 08:02:32|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:02:32|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:257:get_condition_with_negative_prompt] Getting uncondition with negative prompt without image context +[06-26 08:02:32|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:02:50|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:07:56|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 2/2 +[06-26 08:07:56|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 643774 +[06-26 08:07:56|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 1 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 08:07:56|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:255:get_condition_with_negative_prompt] Getting condition with image context +[06-26 08:07:56|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:07:56|INFO|cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/defaults/conditioner.py:257:get_condition_with_negative_prompt] Getting uncondition with negative prompt without image context +[06-26 08:07:56|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:07:58|WARNING|cosmos_transfer2/_src/transfer2/networks/siglip2_image_context.py:83:forward] No image_context provided, using zero latents +[06-26 08:13:00|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:736:generate_img2world] Average time per chunk: 312.4418000805017 +[06-26 08:13:00|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:13:01|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color_control_edge.mp4 +[06-26 08:13:01|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 08:13:01|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 08:13:02|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.mp4 diff --git a/multicontrol_cookbook_grid/sunny/car_color/color/spec.json b/multicontrol_cookbook_grid/sunny/car_color/color/spec.json new file mode 100644 index 0000000000000000000000000000000000000000..7f4274a7347350a084a4ff94b94a8e90dbb2b00d --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/car_color/color/spec.json @@ -0,0 +1,13 @@ +{ + "name": "sunny_car_color_color", + "prompt": "The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. The vehicles moving through the junction are glossy vivid candy-red, their bright paint catching the sunlight and reflecting the sky. A road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, low-rise buildings and sidewalks line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, crisp reflections, detail.", + "video_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4", + "guidance": 3, + "num_steps": 35, + "seed": 2025, + "edge": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4", + "control_weight": 1.0 + }, + "negative_prompt": "The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Dull grey washed-out cars, color bleeding onto road or buildings, melted or deformed vehicle bodies." +} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.json b/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.json new file mode 100644 index 0000000000000000000000000000000000000000..c2484c26b80787b9a10a822a555aa92ca1495b56 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.json @@ -0,0 +1 @@ +{"name":"sunny_car_color_color","prompt_path":null,"prompt":"The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. The vehicles moving through the junction are glossy vivid candy-red, their bright paint catching the sunlight and reflecting the sky. A road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, low-rise buildings and sidewalks line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, crisp reflections, detail.","negative_prompt":"The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Dull grey washed-out cars, color bleeding onto road or buildings, melted or deformed vehicle bodies.","seed":2025,"guidance":3,"video_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4","max_frames":null,"context_frame_index":null,"image_context_path":null,"num_conditional_frames":1,"resolution":"720","sigma_max":null,"num_video_frames_per_chunk":93,"num_steps":35,"show_control_condition":false,"show_input":false,"keep_input_resolution":true,"edge":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4","control_weight":1.0,"mask_path":null,"mask_prompt":null,"preset_edge_threshold":"medium"},"depth":null,"vis":null,"seg":null,"guided_generation_mask":null,"guided_generation_step_threshold":25,"guided_generation_foreground_labels":null} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.mp4 b/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..a5af53f892073a25148fe87e5c3192807bd6416c --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5986d92021638ace09bcdd79b3bbc20be9c28035d34a998d007dc0cc2b0351a1 +size 1831677 diff --git a/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.txt b/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.txt new file mode 100644 index 0000000000000000000000000000000000000000..fbf2a8dffc052cac0a505f5c4495fd4272edebd0 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color.txt @@ -0,0 +1 @@ +The video shows an urban intersection captured from a vehicle-mounted camera in bright daylight. The vehicles moving through the junction are glossy vivid candy-red, their bright paint catching the sunlight and reflecting the sky. A road network with painted lane markings, stop lines and pedestrian crosswalks runs through the scene, low-rise buildings and sidewalks line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, crisp reflections, detail. \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color_control_edge.mp4 b/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color_control_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..517a976ec0a5d8e4a484eb65a5ac5ad8308f733a --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/car_color/color/sunny_car_color_color_control_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57c1f383a1f61fc52a275e2a3d237bee50a45e7901dc709b401226496891cd0c +size 1406752 diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/config.yaml b/multicontrol_cookbook_grid/sunny/weather/lighting/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..cca5228321ac1059bc29ba3371b19bfe5da9908f --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/config.yaml @@ -0,0 +1,825 @@ +checkpoint: + broadcast_via_filesystem: 'False' + dcp_async_mode_enabled: 'False' + enable_gcs_patch_in_boto3: 'False' + jit: + device: cuda + dtype: bfloat16 + enabled: 'False' + input_shape: null + strict: 'True' + keys_not_to_resume: [] + keys_to_skip_loading: [] + load_ema_to_reg: 'False' + load_from_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + load_path: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 + load_training_state: 'False' + only_load_scheduler_state: 'False' + save_iter: '1000' + save_to_object_store: + bucket: bucket + credentials: credentials/s3_checkpoint.secret + enabled: 'True' + strict_resume: 'False' + type: + _target_: + callbacks: null + disable_async: 'False' + verbose: 'True' +dataloader_train: + dataset: + augmentor_name: video_basic_augmentor_v2_with_control + caption_type: t2w_qwen2p5_7b + control_input_type: edge_vis_depth_segcolor + dataset_resolution_type: gt720p + embedding_type: null + max_fps_thres: '60' + min_fps_thres: '10' + num_video_frames: '93' + resolution: '720' + use_native_fps: 'True' + video_decoder_name: video_naive_bytes + num_workers: '4' +dataloader_val: + _target_: + dataloaders: + image_data: + dataloader: + _target_: + batch_size: '2' + cache_augment_fn: null + cache_replay_name: image_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' + video_data: + dataloader: + _target_: + batch_size: '1' + cache_augment_fn: null + cache_replay_name: video_dataloader + cache_size: '32' + concat_size: '1' + dataset: + _target_: + len_t5: '512' + num_video_frames: '136' + resolution: '512' + t5_dim: '1024' + num_workers: '8' + pin_memory: 'True' + shuffle: 'False' + use_cache: 'False' + webdataset: 'False' + ratio: '1' +defaults: +- _self_ +- data_train: mock +- data_val: mock +- optimizer: fusedadamw +- scheduler: lambdalinear +- model: ddp +- callbacks: basic +- net: null +- conditioner: video_prediction_control_conditioner +- ema: power +- tokenizer: wan2pt1_tokenizer +- checkpoint: s3 +- ckpt_type: dummy +- experiment: null +job: + cluster: null + group: vid2vid_2B_control + name: vid2vid_2B_control_720p_t24_control_layer4_cr1pt1_embedding_rectified_flow + project: cosmos_transfer2 + wandb_mode: online +model: + _recursive_: 'False' + _target_: + config: + base_load_from: null + conditional_frame_timestep: -1.0 + conditional_frames_probs: + 0: 0.4 + 1: 0.4 + 2: 0.2 + conditioner: + _target_: + control_input_depth: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth + output_key: control_input_depth + control_input_depth_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_depth_mask + output_key: control_input_depth_mask + control_input_edge: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge + output_key: control_input_edge + control_input_edge_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_edge_mask + output_key: control_input_edge_mask + control_input_inpaint: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint + output_key: control_input_inpaint + control_input_inpaint_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_inpaint_mask + output_key: control_input_inpaint_mask + control_input_seg: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg + output_key: control_input_seg + control_input_seg_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_seg_mask + output_key: control_input_seg_mask + control_input_vis: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis + output_key: control_input_vis + control_input_vis_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: control_input_vis_mask + output_key: control_input_vis_mask + fps: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: fps + output_key: fps + padding_mask: + _target_: + dropout_rate: '0.0' + dtype: null + input_key: padding_mask + output_key: padding_mask + text: + _target_: + credential_path: credentials/s3_training.secret + dropout_rate: '0.2' + empty_string_embeddings_path: s3://bucket/predict2_assets/reason1_empty_string_embeddings.pt + input_key: + - t5_text_embeddings + use_empty_string: 'False' + use_video_condition: + _target_: + dropout_rate: '0.0' + input_key: fps + output_key: use_video_condition + conditioning_strategy: frame_replace + copy_weight_strategy: first_n + denoise_replace_gt_frames: true + ema: + enabled: false + iteration_shift: 0 + rate: 0.1 + fsdp_shard_size: 1 + high_sigma_ratio: 0.05 + high_sigma_timesteps_max: 1000 + high_sigma_timesteps_min: 980 + hint_keys: edge_vis_depth_seg + init_lora_weights: true + input_caption_key: ai_caption + input_data_key: video + input_image_key: images + lora_alpha: 32 + lora_rank: 32 + lora_target_modules: q_proj,k_proj,v_proj,output_proj,mlp.layer1,mlp.layer2 + max_num_conditional_frames: 2 + min_num_conditional_frames: 0 + net: + _target_: + adaln_lora_dim: '256' + atten_backend: minimal_a2a + concat_padding_mask: 'True' + condition_strategy: spaced + crossattn_emb_channels: '1024' + crossattn_proj_in_channels: '100352' + extra_per_block_abs_pos_emb: 'False' + in_channels: '16' + max_frames: '128' + max_img_h: '240' + max_img_w: '240' + mlp_ratio: '4.0' + model_channels: '2048' + num_blocks: '28' + num_control_branches: '4' + num_heads: '16' + num_max_modalities: '8' + out_channels: '16' + patch_spatial: '2' + patch_temporal: '1' + pos_emb_cls: rope3d + pos_emb_interpolation: crop + pos_emb_learnable: 'True' + rope_enable_fps_modulation: 'False' + rope_h_extrapolation_ratio: '3.0' + rope_t_extrapolation_ratio: '1.0' + rope_w_extrapolation_ratio: '3.0' + sac_config: + every_n_blocks: 1 + mode: predict2_2b_720_aggressive + separate_embedders: 'False' + spatial_compression_factor: '8' + timestep_scale: '0.001' + use_adaln_lora: 'True' + use_after_proj_for_multi_branch: 'False' + use_crossattn_projection: 'True' + use_cuda_graphs: 'False' + use_input_hint_block: 'False' + use_wan_fp32_strategy: 'True' + vace_block_every_n: '7' + vace_has_mask: 'False' + precision: bfloat16 + resolution: '720' + shift: 5 + state_ch: 16 + state_t: 24 + text_encoder_class: reason1p1_7B + text_encoder_config: + ckpt_path: s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/ + compute_online: true + embedding_concat_strategy: full_concat + model_config: + _target_: + model_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenModelConfig + activation_checkpoint: + mode: selective + models: vlm + selective_ac_option: op + add_answer_tag: 'True' + add_cross_attention: 'False' + add_image_start_end_tag: 'False' + add_tile_tag: 'False' + architectures: + - Qwen2_5_VLForConditionalGeneration + attention_dropout: '0.0' + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + aux_loss_coeff: '0.0' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: '151643' + cache_dir: null + checkpoint: + async_mode: disabled + create_seed_checkpoint: false + enable_checkpoint: false + export_dtype: float32 + folder: checkpoint + interval: 500 + interval_type: steps + model_weights_only: false + chunk_size_feed_forward: '0' + ckpt_dir: null + ckpt_path: null + comm: + init_timeout_seconds: 300 + trace_buf_size: 20000 + train_timeout_seconds: 100 + cp_size: null + cross_attention_hidden_size: null + decoder_start_token_id: null + deterministic: 'False' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + encoder_no_repeat_ngram_size: '0' + eos_token_id: '151645' + ep_size: null + experimental: + enable_async_tensor_parallel: false + enable_compiled_autograd: false + pipeline_parallel_degree: 1 + exponential_decay_length_penalty: null + finetuning_task: null + float8: + enable_float8_linear: false + forced_bos_token_id: null + forced_eos_token_id: null + freeze_llm: 'False' + freeze_mm_projector: 'False' + freeze_vision_encoder: 'False' + fsdp_enabled: 'False' + hidden_act: silu + hidden_size: '3584' + id2label: + 0: LABEL_0 + 1: LABEL_1 + image_token_id: '151655' + initializer_range: '0.02' + intermediate_size: '18944' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + loss_per_token: 'True' + max_batch_size: '1' + max_length: '20' + max_position_embeddings: '128000' + max_seq_len: '128000' + max_window_layers: '28' + min_length: '0' + mm_projector: null + model_type: qwen2_5_vl + name_or_path: Qwen/Qwen2.5-VL-7B-Instruct + no_repeat_ngram_size: '0' + num_attention_heads: '28' + num_beam_groups: '1' + num_beams: '1' + num_hidden_layers: '28' + num_key_value_heads: '4' + num_return_sequences: '1' + num_tiles: '1' + optimizer: + early_step_in_backward: false + end_lr: 2.5e-05 + fused: false + init_lr: 1.0e-05 + lr: 0.0003 + lr_multiplier_llm: 1.0 + lr_multiplier_mm_projector: 1.0 + lr_multiplier_vision_encoder: 0.1 + name: AdamW + output_attentions: 'False' + output_hidden_states: 'True' + output_scores: 'False' + pad_token_id: null + precision: bfloat16 + prefix: null + prepend_padding: 'False' + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + rms_norm_eps: 1e-06 + rope_scaling: + mrope_section: + - '16' + - '24' + - '24' + rope_type: default + type: default + rope_theta: '1000000.0' + s3_credential_path: credentials/pbss_dir.secret + seed: '0' + sep_token_id: null + sliding_window: '32768' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'False' + tile_tag_type: space_separated + tokenizer_class: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + training: + compile: false + context_parallel_degree: 1 + data_parallel_replicate_degree: 1 + data_parallel_shard_degree: -1 + disable_loss_parallel: false + enable_cpu_offload: false + fsdp_reshard_after_forward: default + mixed_precision_param: bfloat16 + mixed_precision_reduce: float32 + steps: 400000 + tensor_parallel_degree: 1 + use_cosine_decay: false + use_linear_decay: true + warmup_steps: 1000 + training_seq_len: '4096' + transformers_version: 4.51.0.dev0 + typical_p: '1.0' + use_bfloat16: 'False' + use_cache: 'False' + use_fsdp2: 'True' + use_return_dict: 'True' + use_rope_from_torchtitan: 'False' + use_sliding_window: 'False' + video_token_id: '151656' + vision_config: + _target_: cosmos_transfer2._src.reason1.configs.default.model_config_qwen.QwenVisionConfig + add_cross_attention: 'False' + architectures: null + attn_implementation: flash_attention_2 + attn_implementation_autoset: 'True' + bad_words_ids: null + begin_suppress_tokens: null + bos_token_id: null + chunk_size_feed_forward: '0' + cross_attention_hidden_size: null + decoder_start_token_id: null + depth: '32' + diversity_penalty: '0.0' + do_sample: 'False' + early_stopping: 'False' + embed_dim: null + encoder_no_repeat_ngram_size: '0' + eos_token_id: null + exponential_decay_length_penalty: null + finetuning_task: null + forced_bos_token_id: null + forced_eos_token_id: null + fullatt_block_indexes: + - '7' + - '15' + - '23' + - '31' + hidden_act: silu + hidden_size: '1280' + id2label: + 0: LABEL_0 + 1: LABEL_1 + in_channels: '3' + in_chans: '3' + intermediate_size: '3420' + is_decoder: 'False' + is_encoder_decoder: 'False' + label2id: + LABEL_0: '0' + LABEL_1: '1' + length_penalty: '1.0' + max_length: '20' + min_length: '0' + mlp_ratio: null + model_type: qwen2_5_vl + name_or_path: '' + no_repeat_ngram_size: '0' + num_beam_groups: '1' + num_beams: '1' + num_heads: '16' + num_return_sequences: '1' + out_hidden_size: '3584' + output_attentions: 'False' + output_hidden_states: 'False' + output_scores: 'False' + pad_token_id: null + patch_size: '14' + prefix: null + problem_type: null + pruned_heads: _Nothing.NOTHING + remove_invalid_values: 'False' + repetition_penalty: '1.0' + return_dict: 'True' + return_dict_in_generate: 'False' + sep_token_id: null + spatial_merge_size: '2' + spatial_patch_size: '14' + suppress_tokens: null + task_specific_params: null + temperature: '1.0' + temporal_patch_size: '2' + tf_legacy_loss: 'False' + tie_encoder_decoder: 'False' + tie_word_embeddings: 'True' + tokenizer_class: null + tokens_per_second: '2' + top_k: '50' + top_p: '1.0' + torch_dtype: bfloat16 + torchscript: 'False' + typical_p: '1.0' + use_bfloat16: 'False' + window_size: '112' + vision_encoder: openai/clip-vit-base-patch32 + vision_encoder_config: + depth_init: true + dim: 1024 + ffn_dim_multiplier: null + head_dim: null + hidden_act: null + hidden_dim: 4096 + image_size: 1024 + image_token_id: null + multiple_of: null + n_heads: 16 + n_kv_heads: null + n_layers: 24 + norm_eps: 1.0e-05 + norm_type: rmsnorm + num_channels: 3 + patch_size: 16 + proj_bias: null + qkv_bias: null + rope_theta: 10000.0 + use_cache: false + use_rope_from_torchtitan: false + vision_encoder_in_channels: '3' + vision_end_token_id: '151653' + vision_start_token_id: '151652' + vision_token_id: '151654' + vocab_size: '152064' + z_loss_coeff: '0.0' + tokenizer: + _target_: + cache_dir: null + tokenizer_type: Qwen/Qwen2.5-VL-7B-Instruct + n_layers_per_group: 5 + s3_credential_path: credentials/s3_checkpoint.secret + tokenizer: + _target_: + chunk_duration: '81' + load_mean_std: 'False' + name: wan2pt1_tokenizer + temporal_window: '16' + train_time_distribution: logitnormal + train_time_weight: reweighting + use_dora: false + use_dynamic_shift: false + use_high_sigma_strategy: false + use_kerras_sigma_at_inference: false + use_lora: false + use_reference_image: false + use_torch_compile: false +model_parallel: + _cpu_offloading_context: null + async_tensor_model_parallel_allreduce: false + autocast_dtype: torch.float32 + barrier_with_L1_time: true + batch_p2p_comm: true + batch_p2p_sync: true + bf16: false + context_parallel_size: 8 + cpu_offloading: false + cpu_offloading_activations: true + cpu_offloading_double_buffering: false + cpu_offloading_num_layers: 0 + cpu_offloading_weights: true + cross_entropy_fusion_impl: native + cross_entropy_loss_fusion: false + deallocate_pipeline_outputs: false + defer_embedding_wgrad_compute: false + delay_wgrad_compute: false + deterministic_mode: false + enable_autocast: false + expert_model_parallel_size: 1 + expert_tensor_parallel_size: 1 + finalize_model_grads_func: null + fp16: false + grad_scale_func: null + grad_sync_func: null + gradient_accumulation_fusion: false + hierarchical_context_parallel_sizes: null + microbatch_group_size_per_vp_stage: 1 + moe_extended_tp: false + no_sync_func: null + num_microbatches_with_partial_activation_checkpoints: null + overlap_moe_expert_parallel_comm: false + overlap_p2p_comm: false + overlap_p2p_comm_warmup_flush: false + param_sync_func: null + params_dtype: torch.float32 + perform_initialization: true + pipeline_dtype: null + pipeline_model_parallel_comm_backend: null + pipeline_model_parallel_size: 1 + sequence_parallel: false + tensor_model_parallel_size: 1 + timers: null + tp_comm_atomic_ag: false + tp_comm_atomic_rs: false + tp_comm_bootstrap_backend: nccl + tp_comm_bulk_dgrad: true + tp_comm_bulk_wgrad: true + tp_comm_overlap: false + tp_comm_overlap_ag: true + tp_comm_overlap_disable_fc1: false + tp_comm_overlap_disable_qkv: false + tp_comm_overlap_rs: true + tp_comm_overlap_rs_dgrad: false + tp_comm_split_ag: true + tp_comm_split_rs: true + use_cpu_initialization: false + use_ring_exchange_p2p: false + use_te_rng_tracker: false + variable_seq_lengths: false + virtual_pipeline_model_parallel_size: null + wgrad_deferral_limit: 0 +optimizer: + _target_: + betas: + - '0.9' + - '0.999' + eps: 1e-08 + fused: 'True' + lr: '8.63e-05' + model: null + optim_type: adamw + weight_decay: '0.001' +scheduler: + _target_: + cycle_lengths: + - '100000' + f_max: + - '0.5' + f_min: + - '0.2' + f_start: + - 1e-06 + verbosity_interval: '0' + warm_up_steps: + - '100' +trainer: + callbacks: + compile_tokenizer: + _target_: + compile_after_iterations: '4' + dynamic: 'False' + enabled: 'True' + dataloader_speed: + _target_: + every_n: '200' + save_s3: 'True' + step_size: '1' + device_monitor: + _target_: + every_n: '200' + log_memory_detail: 'True' + save_s3: 'True' + step_size: '1' + upload_every_n_mul: '10' + every_n_sample_ema: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'True' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + every_n_sample_reg: + _target_: + every_n: '5000' + fix_batch_fp: null + fps: '16' + guidance: + - '0' + - '3' + - '7' + is_ema: 'False' + is_sample: 'True' + is_x0: 'False' + n_sample_to_save: '128' + n_viz_sample: '3' + n_x0_level: '4' + num_sampling_step: '35' + save_s3: 'True' + show_all_frames: 'False' + step_size: '1' + use_negative_prompt: 'False' + frame_loss_log: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + grad_clip: + _target_: + clip_norm: '0.1' + force_finite: 'True' + heart_beat: + _target_: + every_n: '10' + save_s3: 'True' + step_size: '1' + update_interval_in_minute: '20' + iter_speed: + _target_: + every_n: '100' + hit_thres: '300' + save_s3: 'True' + save_s3_every_log_n: '10' + load_base_model: + _target_: + config: null + trainer: null + low_prec: + _target_: + config: null + trainer: null + update_iter: '1' + manual_gc: + _target_: + every_n: '200' + warm_up: '5' + wandb: + _target_: + logging_iter_multipler: '1' + save_logging_iter_multipler: '10' + save_s3: 'True' + wandb_10x: + _target_: + logging_iter_multipler: '10' + save_logging_iter_multipler: '1' + save_s3: 'True' + compile_config: + recompile_limit: '8' + use_duck_shape: 'True' + cudnn: + benchmark: 'True' + deterministic: 'False' + ddp: + broadcast_buffers: 'True' + find_unused_parameters: 'False' + static_graph: 'True' + distributed_parallelism: fsdp + grad_accum_iter: '1' + grad_scaler_args: + enabled: 'False' + logging_iter: '200' + max_iter: '100000' + max_val_iter: null + memory_format: torch.preserve_format + profiling: + enable_memory_snapshot: 'False' + enable_profiling: 'False' + profile_freq: '1' + profile_memory: 'False' + record_shape: 'False' + save_s3: 'False' + target_ranks: + - '0' + - '1' + - '2' + - '3' + - '4' + - '5' + - '6' + - '7' + with_modules: 'True' + with_stack: 'True' + run_validation: 'False' + run_validation_on_start: 'False' + seed: '0' + straggler_detection: + analyze_backward: 'True' + analyze_dataloading: 'True' + analyze_forward: 'True' + analyze_optimizer: 'True' + enabled: 'True' + max_diff: '1.5' + profile_freq: '1' + raise_error: 'True' + report_freq: '100' + timeout_period: '999999999' + type: + validation_iter: '100' +upload_reproducible_setup: 'True' diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/console.log b/multicontrol_cookbook_grid/sunny/weather/lighting/console.log new file mode 100644 index 0000000000000000000000000000000000000000..860b99df546269d1268232df6a97fec2fd5ee71c --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/console.log @@ -0,0 +1,12 @@ +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:146:__init__] Found 2 hint keys across all samples +[06-26 07:20:53|WARNING|cosmos_transfer2/inference.py:148:__init__] Loading the multicontrol model. Multicontrol inference is not strictly equal to single control +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/config.yaml +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['sunny_weather_lighting'] +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample sunny_weather_lighting +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.json +[06-26 07:20:53|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None, 'vis': None, 'vis_mask': None, 'vis_mask_prompt': None} +[06-26 07:33:20|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_edge.mp4 +[06-26 07:33:21|INFO|cosmos_transfer2/inference.py:311:_generate_sample] vis control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_vis.mp4 +[06-26 07:33:21|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 07:33:22|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.mp4 diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/debug.log b/multicontrol_cookbook_grid/sunny/weather/lighting/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..19100aa7dda1ee49425b7020c554fb3407377c2a --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/debug.log @@ -0,0 +1,33578 @@ +[06-26 07:17:07|DEBUG|packages/cosmos-oss/cosmos_oss/init.py:176:init_output_dir] Flags(internal=False, training=True, smoke=False, verbose=False, experimental_checkpoints=False) +[06-26 07:17:23|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 07:17:23|DEBUG|cosmos_transfer2/_src/imaginaire/utils/env_parsers/env_parser.py:86:get_val_dict] getting val dict of CredentialEnvParser +[06-26 07:19:11|CRITICAL|cosmos_transfer2/_src/interactive/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 07:19:18|CRITICAL|cosmos_transfer2/_src/predict2/checkpointer/dcp.py:181:] for the back comptiable pytorch! New DefaultLoadPlanner class is created. +[06-26 07:19:23|DEBUG|cosmos_transfer2/inference.py:46:__init__] SetupArguments(output_dir=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting') model='edge' checkpoint_path='s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000' experiment='edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5' config_file='cosmos_transfer2/_src/transfer2/configs/vid2vid_transfer/config.py' context_parallel_size=1 disable_guardrails=True offload_guardrail_models=False keep_going=True profile=False benchmark=False compile_tokenizer= enable_parallel_tokenizer=False parallel_tokenizer_grid=(-1, -1))(['edge', 'vis']) +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1... +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2... +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor... +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v1_with_control... +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control... +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering image_basic_augmentor_with_control... +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_v2_with_control_and_image_context... +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering video_basic_augmentor_with_control_input... +[06-26 07:19:27|INFO|cosmos_transfer2/_src/transfer2/datasets/augmentor_provider.py:35:augmentor_register] registering hdmap_augmentor_for_local_datasets... +[06-26 07:19:28|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] Flash Attention v3 is not supported because the Python package ('flash_attn_3_nv'_) was not found. +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/attention/utils/safe_log.py:35:debug] NATTEN Attention is not supported due to insufficient NATTEN version 0.21.0, expected at least 0.21.5.dev9. +[06-26 07:19:29|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_large_scale +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.exp_small_overfit +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment.experiment_list +[06-26 07:19:29|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:206:import_modules_recursively] Skipping module _submit as it starts with an underscore +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.exp_large_scale +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2._src.transfer2.configs.vid2vid_transfer.experiment_av.experiment_list +[06-26 07:19:29|CRITICAL|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:192:import_all_modules_from_package] Reloading all modules from package cosmos_transfer2.experiments +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.agibot.cosmos_agibot_posttrain_example +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview +[06-26 07:19:29|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.multiview.cosmos_multiview_example +[06-26 07:19:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview +[06-26 07:19:30|DEBUG|cosmos_transfer2/_src/imaginaire/utils/config_helper.py:210:import_modules_recursively] Reloading module cosmos_transfer2.experiments.singleview.cosmos_singleview_example +[06-26 07:19:30|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:91:load_model_from_checkpoint] Overriding config checkpoint path with: s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 07:19:31|INFO|cosmos_transfer2/_src/imaginaire/utils/misc.py:151:set_random_seed] Using random seed 0. +[06-26 07:19:31|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:111:load_model_from_checkpoint] Loading model from s3://bucket/cosmos_transfer2/vid2vid_2B_control/edge_720p_t24or1_spaced_layer4_cr1pt1_sdev2_lowsigma0.05_uniform_hqv3p1_20250714_64N_rectified_flow_refimdrop0pt5/checkpoints/iter_000032000 +[06-26 07:19:31|WARNING|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:132:__init__] DiffusionModel: precision torch.bfloat16 +[06-26 07:19:32|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Wan2.1/vae(685afcaa-4de2-42fe-b7b9-69f7a2dee4d8) +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/tokenizers/wan2pt1.py:669:_video_vae] loading /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Predict2.5-2B/snapshots/f176dc95b4a70f53ce01c4b302851595e7322b00/tokenizer.pth +[06-26 07:19:34|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_tokenizer: 2.76 s +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #0-fps: + ReMapkey + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: fps + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #1-padding_mask: + ReMapkey + input key: padding_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: padding_mask + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #2-text: + TextAttr + input key: ['t5_text_embeddings'] + Param count: 0 + Trainable: None + Dropout rate: 0.2 + Output key: [crossattn_emb] +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #3-use_video_condition: + BooleanFlag + input key: fps + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: use_video_condition + This is a boolean flag +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #4-control_input_edge: + ReMapkey + input key: control_input_edge + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #5-control_input_vis: + ReMapkey + input key: control_input_vis + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #6-control_input_depth: + ReMapkey + input key: control_input_depth + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #7-control_input_seg: + ReMapkey + input key: control_input_seg + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #8-control_input_inpaint: + ReMapkey + input key: control_input_inpaint + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #9-control_input_edge_mask: + ReMapkey + input key: control_input_edge_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_edge_mask + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #10-control_input_vis_mask: + ReMapkey + input key: control_input_vis_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_vis_mask + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #11-control_input_depth_mask: + ReMapkey + input key: control_input_depth_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_depth_mask + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #12-control_input_seg_mask: + ReMapkey + input key: control_input_seg_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_seg_mask + Dtype: None +[06-26 07:19:34|INFO|cosmos_transfer2/_src/predict2/conditioner.py:434:__init__] Initialized embedder #13-control_input_inpaint_mask: + ReMapkey + input key: control_input_inpaint_mask + Param count: 0 + Trainable: None + Dropout rate: 0.0 + Output key: control_input_inpaint_mask + Dtype: None +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1624:build_pos_embed] Building positional embedding with rope3d class, impl +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:886:__init__] Using AdaLN LoRA Flag: True. We enable bias if no AdaLN LoRA for backward compatibility. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:34|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is None and using 16 heads with a dimension of 128. +[06-26 07:19:35|DEBUG|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:432:__init__] Setting up Attention. Query dim is 2048, context_dim is 1024 and using 16 heads with a dimension of 128. +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 28 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 4 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 5 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 6 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 7 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 8 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 9 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 10 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 11 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 12 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 13 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 14 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 15 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 16 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 17 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 18 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 19 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 20 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 21 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 22 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 23 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 24 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 25 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 26 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 27 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1804:enable_selective_checkpoint] Enable selective checkpoint with predict2_2b_720_aggressive, for every 1 blocks. Total blocks: 4 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 0 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 1 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 2 +[06-26 07:19:35|INFO|cosmos_transfer2/_src/predict2/networks/minimal_v4_dit.py:1810:enable_selective_checkpoint] Enable selective checkpoint for block 3 +[06-26 07:19:36|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on meta to cuda and broadcast model states: 1.39 s +[06-26 07:19:36|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model: 2.31 s +[06-26 07:19:36|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on Creating PyTorch model and ema if enabled: 2.31 s +[06-26 07:19:36|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on DiffusionModel: set_up_model: 2.32 s +[06-26 07:19:36|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:569:freeze_base_model] +Freezing base model + +[06-26 07:19:36|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:779:load_base_model] Done loading the base model checkpoint. +[06-26 07:19:36|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 0-th block weight to control net's 0-th block +[06-26 07:19:36|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 1-th block weight to control net's 1-th block +[06-26 07:19:36|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 2-th block weight to control net's 2-th block +[06-26 07:19:36|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:532:copy_weights_to_control_branch] ======Copying base model's 3-th block weight to control net's 3-th block +[06-26 07:19:36|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:74:__init__] Instantiating text encoder model... +[06-26 07:19:36|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint Qwen/Qwen2.5-VL-7B-Instruct(7219c6c7-f878-4137-bbdb-76842ea85e70) +[06-26 07:19:37|INFO|cosmos_transfer2/_src/reason1/tokenizer/processor.py:75:__init__] Successfully loaded processor from local cache +[06-26 07:19:37|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:116:__init__] Setting torch default dtype from torch.float32 to torch.bfloat16 +[06-26 07:19:37|INFO|cosmos_transfer2/_src/reason1/models/vlm_base.py:123:__init__] Reset torch default dtype to torch.float32 +[06-26 07:19:38|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:84:__init__] Loading checkpoint from s3://bucket/cosmos_reasoning1/sft_exp700/sft_exp721-1_qwen7b_tl_721_5vs5_s3_balanced_n32_resume_16k/checkpoints/iter_000016000/model/. +[06-26 07:19:38|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Reason1.1-7B(cb3e3ffa-7b08-4c34-822d-61c7aa31a14f) +[06-26 07:20:11|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:113:__init__] Finished loading checkpoint from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Reason1-7B/snapshots/3210bec0495fdc7a8d3dbb8d58da5711eab4b423. +[06-26 07:20:11|INFO|cosmos_transfer2/_src/predict2/text_encoders/text_encoder.py:116:__init__] Text encoder model instantiated +[06-26 07:20:11|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:76:__init__] MinimalV4LVGControlVaceDiT( + (x_embedder): PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=72, out_features=2048, bias=False) + ) + ) + (pos_embedder): VideoRopePosition3DEmb() + (t_embedder): Sequential( + (0): Timesteps() + (1): TimestepEmbedding( + (linear_1): Linear(in_features=2048, out_features=2048, bias=False) + (activation): SiLU() + (linear_2): Linear(in_features=2048, out_features=6144, bias=False) + ) + ) + (blocks): ModuleList( + (0-27): 28 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlAwareDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + ) + ) + ) + (final_layer): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): CheckpointWrapper( + (_checkpoint_wrapped_module): FinalLayer( + (layer_norm): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (linear): Linear(in_features=2048, out_features=64, bias=False) + (adaln_modulation): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=4096, bias=False) + ) + ) + ) + ) + ) + ) + ) + ) + (t_embedding_norm): RMSNorm() + (crossattn_proj): Sequential( + (0): Linear(in_features=100352, out_features=1024, bias=True) + (1): GELU(approximate='none') + ) + (control_embedder): ModuleList( + (0-3): 4 x PatchEmbed( + (proj): Sequential( + (0): Rearrange('b c (t r) (h m) (w n) -> b t h w (c r m n)', r=1, m=2, n=2) + (1): Linear(in_features=520, out_features=2048, bias=False) + ) + ) + ) + (control_blocks_0): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_1): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_2): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) + (control_blocks_3): ModuleList( + (0): CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (before_proj): Linear(in_features=2048, out_features=2048, bias=True) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + (1-3): 3 x CheckpointWrapper( + (_checkpoint_wrapped_module): ControlEncoderDiTBlock( + (layer_norm_self_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (self_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=2048, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=2048, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_cross_attn): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (cross_attn): Attention( + (q_proj): Linear(in_features=2048, out_features=2048, bias=False) + (q_norm): RMSNorm() + (k_proj): Linear(in_features=1024, out_features=2048, bias=False) + (k_norm): RMSNorm() + (v_proj): Linear(in_features=1024, out_features=2048, bias=False) + (v_norm): Identity() + (output_proj): Linear(in_features=2048, out_features=2048, bias=False) + (output_dropout): Identity() + (attn_op): MinimalA2AAttnOp() + ) + (layer_norm_mlp): LayerNorm((2048,), eps=1e-06, elementwise_affine=False) + (mlp): GPT2FeedForward( + (activation): GELU(approximate='none') + (layer1): Linear(in_features=2048, out_features=8192, bias=False) + (layer2): Linear(in_features=8192, out_features=2048, bias=False) + ) + (adaln_modulation_self_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_cross_attn): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (adaln_modulation_mlp): Sequential( + (0): SiLU() + (1): Linear(in_features=2048, out_features=256, bias=False) + (2): Linear(in_features=256, out_features=6144, bias=False) + ) + (after_proj): Linear(in_features=2048, out_features=2048, bias=True) + ) + ) + ) +) +[06-26 07:20:11|INFO|cosmos_transfer2/_src/imaginaire/utils/timer.py:150:_log] Time spent on instantiate model: 39.91 s +[06-26 07:20:11|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 07:20:11|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:227:load_model_state_dict_from_checkpoint] Loading model cached locally from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 07:20:18|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:18|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'control_blocks.0.self_attn.q_proj.weight', 'control_blocks.0.self_attn.q_norm.weight', 'control_blocks.0.self_attn.k_proj.weight', 'control_blocks.0.self_attn.k_norm.weight', 'control_blocks.0.self_attn.v_proj.weight', 'control_blocks.0.self_attn.output_proj.weight', 'control_blocks.0.cross_attn.q_proj.weight', 'control_blocks.0.cross_attn.q_norm.weight', 'control_blocks.0.cross_attn.k_proj.weight', 'control_blocks.0.cross_attn.k_norm.weight', 'control_blocks.0.cross_attn.v_proj.weight', 'control_blocks.0.cross_attn.output_proj.weight', 'control_blocks.0.cross_attn.k_img.weight', 'control_blocks.0.cross_attn.v_img.weight', 'control_blocks.0.cross_attn.q_img.weight', 'control_blocks.0.cross_attn.q_img_norm.weight', 'control_blocks.0.cross_attn.k_img_norm.weight', 'control_blocks.0.mlp.layer1.weight', 'control_blocks.0.mlp.layer2.weight', 'control_blocks.0.adaln_modulation_self_attn.1.weight', 'control_blocks.0.adaln_modulation_self_attn.2.weight', 'control_blocks.0.adaln_modulation_cross_attn.1.weight', 'control_blocks.0.adaln_modulation_cross_attn.2.weight', 'control_blocks.0.adaln_modulation_mlp.1.weight', 'control_blocks.0.adaln_modulation_mlp.2.weight', 'control_blocks.0.before_proj.weight', 'control_blocks.0.before_proj.bias', 'control_blocks.0.after_proj.weight', 'control_blocks.0.after_proj.bias', 'control_blocks.1.self_attn.q_proj.weight', 'control_blocks.1.self_attn.q_norm.weight', 'control_blocks.1.self_attn.k_proj.weight', 'control_blocks.1.self_attn.k_norm.weight', 'control_blocks.1.self_attn.v_proj.weight', 'control_blocks.1.self_attn.output_proj.weight', 'control_blocks.1.cross_attn.q_proj.weight', 'control_blocks.1.cross_attn.q_norm.weight', 'control_blocks.1.cross_attn.k_proj.weight', 'control_blocks.1.cross_attn.k_norm.weight', 'control_blocks.1.cross_attn.v_proj.weight', 'control_blocks.1.cross_attn.output_proj.weight', 'control_blocks.1.cross_attn.k_img.weight', 'control_blocks.1.cross_attn.v_img.weight', 'control_blocks.1.cross_attn.q_img.weight', 'control_blocks.1.cross_attn.q_img_norm.weight', 'control_blocks.1.cross_attn.k_img_norm.weight', 'control_blocks.1.mlp.layer1.weight', 'control_blocks.1.mlp.layer2.weight', 'control_blocks.1.adaln_modulation_self_attn.1.weight', 'control_blocks.1.adaln_modulation_self_attn.2.weight', 'control_blocks.1.adaln_modulation_cross_attn.1.weight', 'control_blocks.1.adaln_modulation_cross_attn.2.weight', 'control_blocks.1.adaln_modulation_mlp.1.weight', 'control_blocks.1.adaln_modulation_mlp.2.weight', 'control_blocks.1.after_proj.weight', 'control_blocks.1.after_proj.bias', 'control_blocks.2.self_attn.q_proj.weight', 'control_blocks.2.self_attn.q_norm.weight', 'control_blocks.2.self_attn.k_proj.weight', 'control_blocks.2.self_attn.k_norm.weight', 'control_blocks.2.self_attn.v_proj.weight', 'control_blocks.2.self_attn.output_proj.weight', 'control_blocks.2.cross_attn.q_proj.weight', 'control_blocks.2.cross_attn.q_norm.weight', 'control_blocks.2.cross_attn.k_proj.weight', 'control_blocks.2.cross_attn.k_norm.weight', 'control_blocks.2.cross_attn.v_proj.weight', 'control_blocks.2.cross_attn.output_proj.weight', 'control_blocks.2.cross_attn.k_img.weight', 'control_blocks.2.cross_attn.v_img.weight', 'control_blocks.2.cross_attn.q_img.weight', 'control_blocks.2.cross_attn.q_img_norm.weight', 'control_blocks.2.cross_attn.k_img_norm.weight', 'control_blocks.2.mlp.layer1.weight', 'control_blocks.2.mlp.layer2.weight', 'control_blocks.2.adaln_modulation_self_attn.1.weight', 'control_blocks.2.adaln_modulation_self_attn.2.weight', 'control_blocks.2.adaln_modulation_cross_attn.1.weight', 'control_blocks.2.adaln_modulation_cross_attn.2.weight', 'control_blocks.2.adaln_modulation_mlp.1.weight', 'control_blocks.2.adaln_modulation_mlp.2.weight', 'control_blocks.2.after_proj.weight', 'control_blocks.2.after_proj.bias', 'control_blocks.3.self_attn.q_proj.weight', 'control_blocks.3.self_attn.q_norm.weight', 'control_blocks.3.self_attn.k_proj.weight', 'control_blocks.3.self_attn.k_norm.weight', 'control_blocks.3.self_attn.v_proj.weight', 'control_blocks.3.self_attn.output_proj.weight', 'control_blocks.3.cross_attn.q_proj.weight', 'control_blocks.3.cross_attn.q_norm.weight', 'control_blocks.3.cross_attn.k_proj.weight', 'control_blocks.3.cross_attn.k_norm.weight', 'control_blocks.3.cross_attn.v_proj.weight', 'control_blocks.3.cross_attn.output_proj.weight', 'control_blocks.3.cross_attn.k_img.weight', 'control_blocks.3.cross_attn.v_img.weight', 'control_blocks.3.cross_attn.q_img.weight', 'control_blocks.3.cross_attn.q_img_norm.weight', 'control_blocks.3.cross_attn.k_img_norm.weight', 'control_blocks.3.mlp.layer1.weight', 'control_blocks.3.mlp.layer2.weight', 'control_blocks.3.adaln_modulation_self_attn.1.weight', 'control_blocks.3.adaln_modulation_self_attn.2.weight', 'control_blocks.3.adaln_modulation_cross_attn.1.weight', 'control_blocks.3.adaln_modulation_cross_attn.2.weight', 'control_blocks.3.adaln_modulation_mlp.1.weight', 'control_blocks.3.adaln_modulation_mlp.2.weight', 'control_blocks.3.after_proj.weight', 'control_blocks.3.after_proj.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_embedder.proj.1.weight'], incorrect_shapes=[]) +[06-26 07:20:18|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:271:load_model_state_dict_from_checkpoint] Checkpoint weights loaded from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt: None +[06-26 07:20:18|INFO|cosmos_transfer2/_src/predict2/utils/model_loader.py:273:load_model_state_dict_from_checkpoint] Checkpoint weights loaded successfully (strict=True). +[06-26 07:20:18|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/edge(61f5694b-0ad5-4ecd-8ad7-c8545627d125) +[06-26 07:20:18|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 0 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/b67b64abda3801a9aceddbff2bdb86126c06db74/general/edge/61f5694b-0ad5-4ecd-8ad7-c8545627d125_ema_bf16.pt +[06-26 07:20:20|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0148, 0.0025, 0.0065], + [-0.0004, 0.0057, -0.0327, ..., -0.0078, 0.0021, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0136, -0.0035, 0.0283], + ..., + [-0.0028, 0.0125, 0.0248, ..., 0.0437, -0.0069, -0.0192], + [ 0.0277, 0.0146, 0.0031, ..., -0.0028, -0.0090, 0.0146], + [ 0.0437, 0.0366, 0.0082, ..., 0.0029, -0.0126, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0044, -0.0310, 0.0300, ..., -0.0064, -0.0045, -0.0339], + [-0.0103, -0.0047, 0.0114, ..., 0.0052, 0.0092, 0.0520], + [-0.0061, 0.0131, 0.0398, ..., -0.0096, -0.0179, -0.0056], + ..., + [-0.0048, 0.0175, -0.0101, ..., 0.0225, 0.0043, 0.0042], + [-0.0209, -0.0081, 0.0089, ..., -0.0308, 0.0108, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0055, 0.0354, -0.0082, ..., 0.0120, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0013, ..., 0.0376, 0.0532, 0.0036], + [ 0.0175, -0.0347, 0.0013, ..., -0.0066, 0.0053, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0154, 0.0262, -0.0112], + [-0.0006, 0.0244, 0.0356, ..., -0.0036, 0.0170, 0.0118], + [ 0.0486, 0.0175, 0.0042, ..., 0.0036, -0.0026, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0114, ..., -0.0070, -0.0337, 0.0077], + [ 0.0491, 0.0413, 0.0251, ..., -0.0030, 0.0393, 0.0042], + [-0.0162, -0.0019, 0.0140, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0038, 0.0261, 0.0039, ..., -0.0040, 0.0347, 0.0009], + [-0.0254, 0.0063, 0.0019, ..., 0.0093, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0121, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0035, -0.0270, -0.0166, ..., 0.0115, -0.0298, 0.0532], + [ 0.0017, 0.0040, -0.0131, ..., -0.0161, 0.0146, 0.0308], + [ 0.0146, 0.0087, 0.0143, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0064, -0.0630, -0.0048, ..., 0.0074, -0.0183, 0.0217], + [ 0.0023, -0.0035, 0.0258, ..., -0.0486, -0.0349, -0.0059], + [ 0.0182, -0.0076, 0.0197, ..., 0.0142, -0.0226, -0.0212]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0051, -0.0012, -0.0011, ..., -0.0005, -0.0018, -0.0120], + [-0.0153, 0.0255, 0.0432, ..., -0.0045, 0.0049, -0.0069], + [-0.0154, -0.0206, -0.0245, ..., -0.0119, -0.0325, -0.0013], + ..., + [-0.0034, 0.0039, 0.0417, ..., 0.0280, 0.0036, -0.0047], + [-0.0054, 0.0041, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0055, -0.0024, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 1.5503e-02, -6.5613e-03, -4.8096e-02, ..., 4.4922e-02, + 2.1240e-02, 1.2756e-02], + [ 1.5381e-02, -2.2705e-02, 2.1973e-02, ..., -2.2095e-02, + 3.5858e-03, 6.9885e-03], + [-1.5259e-02, -1.0071e-02, 4.2969e-02, ..., -9.5825e-03, + 3.2715e-02, -1.1169e-02], + ..., + [ 3.1006e-02, -3.7598e-02, -4.3457e-02, ..., -2.2461e-02, + 2.5513e-02, 1.3428e-02], + [-3.8862e-05, 1.6602e-02, -1.6174e-03, ..., -4.0283e-03, + 6.7139e-03, -9.6436e-03], + [-1.0864e-02, -3.4637e-03, 1.1780e-02, ..., 4.6997e-03, + -3.7994e-03, -7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[ 0.0049, -0.0386, -0.0112, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0079, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0127, 0.0032, -0.0114, ..., 0.0214, 0.0110, -0.0045], + ..., + [-0.0155, -0.0172, -0.0136, ..., 0.0095, 0.0098, 0.0184], + [-0.0151, -0.0356, 0.0200, ..., -0.0060, -0.0112, 0.0332], + [-0.0396, -0.0120, -0.0089, ..., -0.0284, -0.0140, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 2.7710e-02, 3.1250e-02, 1.3306e-02, ..., -1.9531e-02, + 1.1414e-02, -1.4954e-02], + [ 1.9165e-02, 9.2773e-03, -7.5378e-03, ..., -1.7853e-03, + -1.2695e-02, -1.9775e-02], + [-2.5024e-02, 1.1108e-02, -1.3733e-02, ..., 1.0132e-02, + 1.7212e-02, 7.0496e-03], + ..., + [ 4.6631e-02, 6.4392e-03, -1.5869e-03, ..., -4.0283e-03, + 2.3926e-02, 2.5330e-03], + [ 5.4688e-02, -1.8677e-02, -1.1108e-02, ..., -3.7537e-03, + 3.1128e-02, -5.4626e-03], + [ 8.7738e-05, 1.0742e-02, -9.6436e-03, ..., 2.0020e-02, + 1.6968e-02, 1.4038e-02]], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[-5.6028e-05, 2.3651e-03, 2.8564e-02, ..., -1.0376e-02, + 4.5471e-03, -1.8188e-02], + [ 6.2866e-03, 1.3428e-02, -3.1494e-02, ..., -4.5654e-02, + -5.0049e-03, 1.0376e-02], + [ 4.3945e-03, 2.6123e-02, 2.3926e-02, ..., 5.0049e-03, + 6.6833e-03, 4.6143e-02], + ..., + [ 1.6968e-02, -2.5391e-02, -4.3945e-02, ..., 1.5259e-02, + 4.9133e-03, -7.4768e-03], + [ 5.7678e-03, -3.7109e-02, 3.5645e-02, ..., -8.3008e-03, + 4.1016e-02, -1.7822e-02], + [ 1.7822e-02, -1.4648e-02, 5.6152e-02, ..., -2.9053e-02, + -2.6733e-02, 2.9663e-02]], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0067, 0.0208, 0.0278, ..., 0.0133, 0.0156, 0.0074], + [-0.0063, 0.0040, 0.0014, ..., 0.0135, 0.0040, 0.0045], + [ 0.0058, 0.0041, 0.0243, ..., -0.0123, -0.0049, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0064, ..., 0.0250, -0.0107, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., 0.0048, -0.0295, 0.0400], + [-0.0116, 0.0138, 0.0305, ..., -0.0103, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0054, 0.0381, -0.0239, ..., 0.0219, 0.0483, 0.0339], + [-0.0143, -0.0033, 0.0288, ..., 0.0135, 0.0583, 0.0049], + [-0.0280, 0.0154, -0.0386, ..., -0.0444, -0.0092, -0.0026], + ..., + [-0.0038, 0.0157, 0.0217, ..., 0.0048, 0.0114, 0.0154], + [ 0.0081, 0.0195, -0.0283, ..., -0.0060, -0.0021, -0.0342], + [-0.0166, -0.0195, 0.0027, ..., -0.0066, 0.0165, 0.0130]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0138, -0.0146, 0.0459, ..., -0.0028, -0.0320, 0.0277], + [-0.0026, 0.0068, 0.0270, ..., -0.0220, -0.0183, -0.0271], + [ 0.0221, 0.0219, -0.0201, ..., 0.0060, 0.0064, -0.0031], + ..., + [-0.0258, 0.0016, 0.0215, ..., 0.0038, 0.0237, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0055], + [-0.0006, -0.0080, -0.0120, ..., -0.0334, 0.0047, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0147, -0.0454, ..., -0.0132, -0.0125, 0.0400], + [ 0.0295, -0.0187, -0.0110, ..., -0.0283, 0.0047, -0.0186], + [-0.0007, 0.0547, 0.0024, ..., 0.0116, 0.0400, -0.0214], + ..., + [ 0.0280, -0.0058, -0.0266, ..., -0.0084, -0.0287, 0.0150], + [ 0.0090, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0095, 0.0102, ..., -0.0070, 0.0703, -0.0043]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0205, 0.0087, -0.0315, ..., 0.0081, 0.0254, 0.0359], + [ 0.0097, -0.0032, 0.0038, ..., -0.0120, -0.0233, -0.0048], + [ 0.0051, 0.0040, -0.0121, ..., -0.0117, -0.0070, 0.0086], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0040], + [ 0.0339, -0.0064, 0.0223, ..., 0.0410, -0.0017, 0.0136], + [-0.0046, -0.0439, -0.0486, ..., 0.0145, -0.0130, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0046, 0.0139, -0.0030, ..., -0.0081, -0.0033, -0.0229], + [-0.0289, 0.0231, 0.0077, ..., -0.0135, -0.0024, 0.0222], + [ 0.0083, 0.0295, -0.0189, ..., -0.0168, 0.0311, -0.0261], + ..., + [ 0.0233, -0.0045, -0.0036, ..., 0.0032, 0.0278, 0.0049], + [-0.0106, -0.0038, 0.0140, ..., -0.0175, 0.0444, 0.0425], + [ 0.0069, -0.0078, 0.0102, ..., 0.0066, -0.0344, 0.0229]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[-0.0050, 0.0128, 0.0190, ..., -0.0093, -0.0198, 0.0149], + [-0.0581, 0.0294, -0.0047, ..., -0.0240, 0.0043, 0.0300], + [ 0.0085, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0035, -0.0289, 0.0537, ..., 0.0058, -0.0074, -0.0099], + [-0.0029, -0.0069, 0.0210, ..., -0.0021, 0.0091, 0.0186], + [-0.0086, -0.0349, -0.0173, ..., -0.0400, 0.0052, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0177, -0.0261, -0.0112, ..., 0.0110, -0.0130, -0.0405], + [ 0.0058, -0.0086, 0.0266, ..., -0.0400, 0.0073, 0.0391], + [-0.0013, 0.0359, 0.0284, ..., 0.0286, -0.0234, 0.0217], + ..., + [ 0.0017, 0.0123, 0.0132, ..., -0.0138, 0.0032, 0.0096], + [ 0.0299, -0.0167, -0.0022, ..., -0.0063, 0.0043, -0.0206], + [ 0.0074, -0.0076, -0.0043, ..., -0.0047, -0.0067, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, -0.0032, 0.0121, ..., -0.0075, -0.0208, -0.0058], + [-0.0092, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0229], + [ 0.0132, -0.0278, 0.0115, ..., 0.0396, -0.0082, -0.0074], + ..., + [-0.0015, 0.0018, -0.0309, ..., 0.0344, 0.0073, -0.0126], + [-0.0236, -0.0102, -0.0056, ..., 0.0332, -0.0117, -0.0171], + [ 0.0231, 0.0405, 0.0103, ..., -0.0278, -0.0139, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[-0.0046, 0.0166, 0.0078, ..., 0.0273, -0.0352, 0.0153], + [-0.0510, -0.0070, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0077, ..., -0.0388, -0.0047, 0.0398], + ..., + [-0.0160, -0.0195, -0.0051, ..., -0.0040, -0.0055, -0.0035], + [-0.0187, -0.0040, 0.0097, ..., 0.0287, -0.0122, -0.0099], + [-0.0181, -0.0013, 0.0247, ..., -0.0145, 0.0072, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0078, 0.0092, 0.0140, ..., -0.0121, 0.0045, -0.0253], + [-0.0164, 0.0330, -0.0171, ..., -0.0033, -0.0403, -0.0111], + [ 0.0085, 0.0347, 0.0275, ..., 0.0151, 0.0069, 0.0080], + ..., + [ 0.0275, -0.0168, -0.0106, ..., -0.0099, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0056, ..., -0.0023, -0.0022, -0.0243], + [-0.0060, 0.0581, 0.0061, ..., 0.0148, 0.0239, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0121, -0.0410, 0.0128, ..., -0.0254, -0.0036, -0.0173], + [ 0.0476, 0.0315, 0.0141, ..., -0.0154, -0.0073, 0.0048], + [-0.0092, -0.0270, 0.0092, ..., 0.0332, 0.0150, -0.0073], + ..., + [-0.0076, 0.0039, 0.0162, ..., -0.0177, 0.0171, -0.0133], + [ 0.0044, -0.0187, -0.0051, ..., 0.0248, -0.0145, -0.0073], + [-0.0143, 0.0034, 0.0149, ..., -0.0047, -0.0006, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0049, 0.0097, ..., 0.0255, -0.0253, -0.0136], + [-0.0038, -0.0141, -0.0160, ..., -0.0109, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0117, 0.0056, 0.0322, ..., -0.0183, -0.0041, -0.0062], + [ 0.0045, -0.0054, 0.0164, ..., 0.0036, 0.0255, 0.0024], + [ 0.0270, 0.0084, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0077, ..., -0.0532, 0.0106, 0.0036], + [-0.0240, -0.0215, -0.0104, ..., -0.0063, 0.0097, -0.0168], + [ 0.0075, 0.0064, -0.0055, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0059, 0.0172, -0.0106, ..., -0.0300, -0.0005, 0.0435], + [ 0.0053, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0061], + [-0.0167, 0.0128, 0.0101, ..., -0.0247, 0.0119, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0065, -0.0079, 0.0317, ..., 0.0200, -0.0105, -0.0376], + [ 0.0356, 0.0378, 0.0129, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., 0.0076, 0.0065, -0.0322], + ..., + [ 0.0065, 0.0107, -0.0029, ..., 0.0366, -0.0206, 0.0374], + [-0.0034, -0.0079, -0.0044, ..., -0.0095, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0048, ..., -0.0166, -0.0042, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0078, -0.0096, -0.0344, ..., -0.0003, 0.0347, -0.0181], + [ 0.0267, -0.0366, -0.0089, ..., 0.0217, 0.0349, -0.0070], + [-0.0098, -0.0286, 0.0062, ..., 0.0060, -0.0282, 0.0047], + ..., + [-0.0124, -0.0039, 0.0065, ..., -0.0084, 0.0165, 0.0074], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0134], + [-0.0140, 0.0049, -0.0170, ..., -0.0126, 0.0060, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0033, 0.0051, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0120, 0.0374, ..., -0.0153, -0.0188, -0.0137], + ..., + [ 0.0262, 0.0052, -0.0065, ..., 0.0007, 0.0038, -0.0165], + [-0.0222, -0.0155, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0125, -0.0148, 0.0038, ..., 0.0018, -0.0234, 0.0017]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0102], + [ 0.0249, 0.0205, 0.0054, ..., -0.0046, -0.0093, -0.0106], + [-0.0220, -0.0004, -0.0037, ..., 0.0306, 0.0381, -0.0044], + ..., + [ 0.0200, -0.0050, 0.0170, ..., 0.0116, 0.0320, -0.0157], + [ 0.0349, 0.0119, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0049, 0.0256, 0.0045, ..., 0.0205, 0.0002, -0.0105]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0227, 0.0065, 0.0199, ..., -0.0057, 0.0276, -0.0312], + [ 0.0178, -0.0059, 0.0103, ..., 0.0112, -0.0398, 0.0293], + [ 0.0430, -0.0130, 0.0072, ..., 0.0040, -0.0069, -0.0029], + ..., + [ 0.0192, -0.0112, -0.0112, ..., 0.0027, 0.0330, 0.0137], + [-0.0125, -0.0542, -0.0205, ..., 0.0044, -0.0159, 0.0320], + [ 0.0303, 0.0047, -0.0425, ..., 0.0032, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0060, -0.0315, ..., 0.0525, 0.0041, -0.0349], + [-0.0036, -0.0026, -0.0156, ..., 0.0048, -0.0240, 0.0305], + [-0.0256, 0.0064, 0.0065, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0065, 0.0231, 0.0209, ..., 0.0076, -0.0041, 0.0461], + [-0.0063, -0.0079, -0.0019, ..., -0.0217, 0.0347, -0.0040], + [ 0.0050, 0.0234, 0.0031, ..., 0.0261, -0.0052, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0052, 0.0046], + [-0.0183, -0.0211, -0.0052, ..., -0.0149, 0.0139, -0.0162], + [-0.0078, -0.0160, 0.0079, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0096, 0.0073, ..., 0.0171, 0.0036, -0.0033], + [-0.0144, 0.0260, -0.0087, ..., 0.0041, 0.0106, -0.0166], + [ 0.0098, -0.0140, -0.0014, ..., 0.0057, 0.0054, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0078, 0.0134, -0.0064, ..., -0.0278, 0.0240, -0.0123], + [-0.0069, 0.0188, -0.0160, ..., -0.0317, -0.0064, 0.0181], + [-0.0042, 0.0310, -0.0085, ..., -0.0226, 0.0017, 0.0334], + ..., + [-0.0081, 0.0186, -0.0073, ..., 0.0228, 0.0047, 0.0064], + [ 0.0271, 0.0118, 0.0354, ..., -0.0075, 0.0198, 0.0150], + [ 0.0282, 0.0171, 0.0317, ..., -0.0109, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-1.1475e-02, -2.6489e-02, -1.0986e-02, ..., 4.8523e-03, + 3.7231e-03, 1.3428e-02], + [-3.3203e-02, -4.8523e-03, 3.0640e-02, ..., 3.4668e-02, + -3.6133e-02, 8.4839e-03], + [-5.8651e-05, -1.9043e-02, -3.6774e-03, ..., 1.6602e-02, + -1.3550e-02, 1.3733e-02], + ..., + [-2.0020e-02, -2.9144e-03, 2.4414e-04, ..., -3.3936e-02, + 3.6133e-02, 3.0518e-03], + [ 6.6528e-03, -2.5024e-02, 1.5564e-02, ..., 2.2430e-03, + -1.6235e-02, 2.1973e-02], + [-7.4463e-03, 3.8452e-03, 2.0508e-02, ..., -6.5002e-03, + -8.3008e-03, 4.3457e-02]], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0047, 0.0240, -0.0217, ..., 0.0172, 0.0046, 0.0114], + [-0.0386, 0.0106, 0.0078, ..., -0.0060, 0.0322, 0.0086], + [-0.0427, 0.0011, -0.0045, ..., 0.0076, 0.0112, -0.0137], + ..., + [-0.0087, -0.0037, -0.0264, ..., -0.0586, -0.0630, -0.0131], + [-0.0006, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0107, -0.0610, 0.0162, ..., -0.0291, -0.0166, 0.0029]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0117, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0112], + [ 0.0111, -0.0098, -0.0327, ..., 0.0447, 0.0076, 0.0059], + [ 0.0042, -0.0175, 0.0076, ..., -0.0069, -0.0037, -0.0060], + ..., + [ 0.0295, 0.0042, -0.0102, ..., 0.0099, -0.0160, -0.0286], + [ 0.0082, 0.0253, -0.0177, ..., 0.0095, 0.0064, 0.0078], + [-0.0132, -0.0048, 0.0139, ..., -0.0125, 0.0277, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0129, 0.0040, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0050, 0.0234, 0.0069, ..., 0.0464, 0.0056, 0.0309], + ..., + [-0.0118, -0.0175, 0.0327, ..., 0.0132, 0.0051, 0.0141], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, 0.0040, -0.0444, ..., 0.0060, -0.0283, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0081, -0.0100, 0.0146, ..., -0.0742, -0.0074, 0.0161], + [-0.0064, -0.0132, -0.0071, ..., 0.0053, 0.0042, 0.0238], + [ 0.0189, 0.0221, -0.0087, ..., 0.0033, 0.0408, 0.0064], + ..., + [ 0.0047, 0.0247, 0.0396, ..., 0.0032, -0.0304, -0.0183], + [ 0.0299, -0.0113, -0.0417, ..., -0.0052, -0.0376, 0.0032], + [-0.0210, 0.0168, -0.0039, ..., 0.0052, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0056, 0.0242, -0.0079, ..., -0.0245, 0.0050, -0.0057], + [ 0.0190, -0.0123, 0.0332, ..., -0.0143, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0053, ..., 0.0056, -0.0091, 0.0182], + ..., + [-0.0430, -0.0403, -0.0086, ..., 0.0134, -0.0140, 0.0060], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0089, 0.0378], + [ 0.0364, 0.0486, -0.0120, ..., -0.0060, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0094, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0078, ..., 0.0101, -0.0223, -0.0052], + ..., + [ 0.0192, 0.0292, -0.0022, ..., 0.0176, -0.0197, 0.0076], + [ 0.0265, 0.0064, 0.0057, ..., 0.0286, 0.0046, -0.0267], + [-0.0184, -0.0088, -0.0045, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0083, -0.0159, 0.0082, ..., 0.0198, 0.0044, 0.0277], + [ 0.0172, 0.0037, -0.0053, ..., 0.0054, -0.0044, 0.0074], + [-0.0019, 0.0184, 0.0113, ..., 0.0099, -0.0054, -0.0077], + ..., + [ 0.0061, -0.0111, -0.0028, ..., -0.0095, -0.0038, 0.0123], + [-0.0427, -0.0391, -0.0193, ..., 0.0096, -0.0238, -0.0078], + [ 0.0040, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0217, -0.0226, -0.0349, ..., -0.0079, -0.0161, 0.0127], + [-0.0049, 0.0008, 0.0312, ..., 0.0055, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0076, 0.0004, 0.0278], + ..., + [ 0.0170, -0.0082, 0.0471, ..., 0.0018, 0.0083, -0.0049], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0079, -0.0203], + [ 0.0033, 0.0081, -0.0077, ..., -0.0098, -0.0076, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[-0.0044, -0.0151, 0.0284, ..., -0.0109, 0.0140, -0.0002], + [ 0.0156, 0.0332, 0.0337, ..., 0.0223, -0.0165, -0.0015], + [-0.0071, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0068, 0.0062, 0.0280, ..., -0.0066, -0.0072, 0.0137], + [-0.0073, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0204], + [-0.0206, 0.0159, -0.0091, ..., 0.0045, 0.0167, -0.0086]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0055, ..., -0.0238, -0.0102, 0.0119], + [-0.0250, -0.0162, 0.0019, ..., -0.0084, 0.0035, -0.0369], + [ 0.0004, 0.0141, 0.0061, ..., -0.0452, -0.0056, 0.0121], + ..., + [ 0.0147, -0.0140, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0099, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0141, -0.0048, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[ 0.0013, -0.0327, -0.0190, ..., 0.0053, -0.0018, -0.0209], + [ 0.0713, 0.0150, -0.0026, ..., 0.0073, 0.0186, -0.0029], + [-0.0227, -0.0090, 0.0188, ..., -0.0125, 0.0193, -0.0049], + ..., + [ 0.0117, 0.0464, 0.0079, ..., 0.0248, 0.0168, -0.0038], + [ 0.0381, 0.0045, 0.0067, ..., 0.0048, -0.0110, 0.0035], + [ 0.0292, 0.0055, 0.0209, ..., 0.0044, 0.0234, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0063, 0.0212, ..., 0.0305, 0.0137, 0.0104], + [-0.0072, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0050], + ..., + [ 0.0082, -0.0275, 0.0225, ..., 0.0051, -0.0189, -0.0104], + [-0.0125, -0.0051, -0.0041, ..., 0.0046, -0.0048, 0.0211], + [ 0.0046, -0.0064, -0.0060, ..., -0.0109, 0.0264, 0.0047]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[-0.0002, -0.0076, -0.0109, ..., 0.0186, -0.0216, 0.0062], + [ 0.0045, 0.0123, -0.0231, ..., 0.0125, -0.0114, -0.0183], + [ 0.0107, 0.0103, -0.0038, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0121, 0.0146, ..., 0.0540, 0.0415, 0.0096], + [ 0.0061, 0.0084, -0.0143, ..., 0.0060, -0.0189, -0.0237], + [ 0.0048, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0067]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0019, 0.0359, ..., -0.0083, -0.0123, -0.0050], + [ 0.0190, -0.0044, 0.0298, ..., 0.0127, 0.0042, -0.0200], + [ 0.0114, -0.0112, 0.0177, ..., -0.0248, 0.0229, 0.0025], + ..., + [-0.0082, -0.0049, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0042, -0.0291, ..., 0.0206, 0.0304, -0.0089], + [-0.0081, 0.0245, -0.0066, ..., -0.0413, 0.0216, -0.0117]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0078, -0.0157, 0.0077, ..., -0.0049, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0064], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, 0.0038], + ..., + [-0.0036, 0.0003, 0.0258, ..., -0.0310, 0.0091, -0.0361], + [-0.0078, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [ 0.0067, 0.0157, -0.0070, ..., 0.0040, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0077, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0106, ..., 0.0103, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0041, ..., 0.0129, -0.0046, 0.0085], + ..., + [-0.0225, 0.0337, -0.0085, ..., -0.0177, 0.0542, -0.0249], + [ 0.0013, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0128, 0.0206, ..., 0.0071, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0095, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0052, 0.0063, ..., 0.0039, 0.0113, 0.0598], + [-0.0068, 0.0267, 0.0054, ..., 0.0284, -0.0060, -0.0364], + ..., + [ 0.0044, 0.0040, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0046, -0.0089, ..., 0.0211, 0.0135, -0.0075], + [ 0.0101, 0.0236, -0.0206, ..., -0.0135, -0.0048, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0067, 0.0294, ..., 0.0050, 0.0270, 0.0063], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0065, -0.0168], + [ 0.0201, -0.0120, 0.0126, ..., -0.0381, -0.0247, -0.0076], + ..., + [ 0.0275, -0.0182, -0.0183, ..., 0.0464, 0.0091, -0.0077], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0067, -0.0466], + [ 0.0327, 0.0284, -0.0030, ..., 0.0143, 0.0142, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, -0.0027, -0.0167, ..., -0.0039, 0.0170, 0.0107], + [-0.0002, -0.0002, 0.0295, ..., -0.0172, 0.0048, 0.0135], + [ 0.0193, 0.0236, -0.0123, ..., -0.0007, -0.0369, 0.0239], + ..., + [ 0.0150, -0.0413, -0.0126, ..., -0.0354, 0.0052, -0.0320], + [-0.0265, -0.0078, -0.0327, ..., -0.0070, 0.0077, 0.0074], + [ 0.0299, 0.0356, -0.0192, ..., 0.0059, -0.0052, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0013, 0.0041, 0.0317, ..., 0.0042, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0076, -0.0093], + [-0.0166, 0.0072, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0019, 0.0260, ..., 0.0159, 0.0048, -0.0474], + [-0.0056, 0.0080, 0.0077, ..., 0.0146, 0.0557, 0.0048], + [-0.0203, 0.0273, 0.0047, ..., -0.0330, 0.0065, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0058, ..., -0.0014, -0.0093, -0.0277], + [ 0.0060, -0.0078, -0.0062, ..., 0.0036, 0.0140, -0.0161], + [-0.0052, 0.0168, -0.0059, ..., -0.0092, 0.0124, 0.0043], + ..., + [ 0.0398, 0.0452, 0.0251, ..., 0.0039, -0.0330, 0.0051], + [-0.0620, -0.0039, 0.0371, ..., 0.0410, 0.0061, -0.0173], + [ 0.0127, 0.0074, -0.0270, ..., 0.0081, 0.0388, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0040, 0.0245], + [-0.0449, 0.0025, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0064, ..., 0.0046, 0.0288, 0.0125], + [ 0.0330, 0.0038, 0.0057, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0069, 0.0045, ..., -0.0077, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0070, 0.0038, 0.0527], + [-0.0051, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0095, -0.0096, 0.0327], + [ 0.0146, 0.0116, -0.0030, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0129, 0.0281, ..., -0.0242, -0.0126, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0155, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0041], + [-0.0048, -0.0045, -0.0286, ..., 0.0050, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0057, ..., -0.0019, 0.0082, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0096, 0.0150], + [-0.0071, 0.0029, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0063, ..., 0.0092, -0.0114, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0070, -0.0033, ..., -0.0220, -0.0145, -0.0046], + [-0.0166, 0.0189, -0.0058, ..., 0.0096, 0.0033, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0067, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0066, ..., 0.0024, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0069, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0121, 0.0289, 0.0126, ..., 0.0134, 0.0066, -0.0103], + [ 0.0188, -0.0197, -0.0022, ..., 0.0284, -0.0342, -0.0058], + [-0.0525, -0.0214, 0.0154, ..., -0.0270, 0.0081, 0.0254], + ..., + [-0.0052, 0.0087, 0.0281, ..., 0.0129, -0.0042, 0.0049], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0037, -0.0054], + [-0.0273, -0.0747, 0.0208, ..., -0.0042, -0.0052, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 9.1171e-04, 1.4954e-02, 3.3447e-02, ..., 2.0508e-02, + -2.7344e-02, -3.7537e-03], + [-1.6479e-02, 1.1108e-02, -1.3489e-02, ..., 1.0620e-02, + -7.2327e-03, 2.1851e-02], + [ 5.0659e-03, 2.2705e-02, -5.6458e-03, ..., 1.2817e-02, + -7.1106e-03, 1.4954e-02], + ..., + [-3.4668e-02, -1.4709e-02, 6.8359e-03, ..., 5.5847e-03, + -5.0659e-03, -9.2773e-03], + [-6.5613e-03, 3.9795e-02, -3.4424e-02, ..., 2.4536e-02, + 6.3477e-02, -1.1780e-02], + [-5.3024e-04, -5.0049e-02, -2.3193e-02, ..., 3.1982e-02, + -7.5340e-05, -5.5847e-03]], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0154, ..., -0.0344, 0.0067, -0.0364], + [ 0.0220, -0.0030, -0.0131, ..., 0.0078, 0.0101, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, 0.0072, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0073, 0.0192, -0.0393], + [-0.0105, -0.0244, 0.0275, ..., -0.0052, 0.0042, 0.0081], + [-0.0170, -0.0427, 0.0142, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0046, -0.0378, -0.0359, ..., 0.0552, -0.0078, -0.0160], + [ 0.0045, 0.0140, 0.0133, ..., 0.0050, -0.0153, 0.0354], + [ 0.0095, -0.0042, 0.0045, ..., -0.0047, -0.0249, -0.0136], + ..., + [ 0.0079, -0.0125, 0.0051, ..., -0.0095, -0.0084, 0.0052], + [-0.0060, 0.0439, 0.0166, ..., 0.0248, 0.0132, -0.0187], + [ 0.0175, 0.0063, 0.0044, ..., 0.0410, -0.0096, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0105, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0145, 0.0025, ..., 0.0065, -0.0244, 0.0356], + [ 0.0079, 0.0165, -0.0200, ..., -0.0080, 0.0237, -0.0081], + ..., + [-0.0197, -0.0022, 0.0386, ..., 0.0181, 0.0113, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, -0.0041, -0.0374], + [-0.0086, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0075, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0137, 0.0115, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, 0.0013, -0.0066, ..., 0.0050, -0.0045, 0.0337], + ..., + [ 0.0422, 0.0054, 0.0352, ..., -0.0304, -0.0035, -0.0204], + [ 0.0302, 0.0156, 0.0100, ..., -0.0049, -0.0254, -0.0245], + [-0.0067, -0.0064, 0.0078, ..., -0.0078, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0044, 0.0266, ..., 0.0154, 0.0187, 0.0049], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0096], + [-0.0189, 0.0205, 0.0044, ..., -0.0172, -0.0090, 0.0231], + ..., + [-0.0295, -0.0137, 0.0045, ..., 0.0039, -0.0050, 0.0099], + [-0.0052, -0.0140, 0.0317, ..., -0.0153, -0.0459, -0.0065], + [-0.0041, -0.0051, -0.0259, ..., -0.0040, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0059, -0.0140, ..., 0.0242, 0.0055, -0.0386], + [-0.0141, 0.0060, -0.0097, ..., -0.0369, -0.0155, 0.0157], + [-0.0267, -0.0041, -0.0065, ..., -0.0127, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0029, 0.0144, ..., 0.0115, 0.0040, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., 0.0045, -0.0038, -0.0137], + [-0.0219, -0.0396, -0.0010, ..., -0.0077, -0.0033, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0051, 0.0168, 0.0194, ..., 0.0349, -0.0420, -0.0048], + [ 0.0044, 0.0273, 0.0146, ..., -0.0184, 0.0011, 0.0156], + [ 0.0145, -0.0031, -0.0070, ..., 0.0245, 0.0359, 0.0061], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0037, 0.0043, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0072, -0.0312, -0.0457], + [ 0.0055, 0.0039, -0.0058, ..., -0.0038, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0047, 0.0121, 0.0038, ..., 0.0124, 0.0006, 0.0067], + [-0.0277, -0.0039, -0.0181, ..., 0.0005, 0.0020, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0038, 0.0513, -0.0113], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0113], + [-0.0282, -0.0040, 0.0121, ..., -0.0057, -0.0039, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, 0.0048, -0.0128, ..., 0.0140, 0.0308, -0.0066], + [ 0.0120, -0.0332, -0.0320, ..., -0.0045, -0.0140, -0.0039], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0088, 0.0295, 0.0505], + [-0.0159, 0.0025, -0.0311, ..., 0.0120, 0.0063, 0.0081], + [ 0.0044, 0.0198, 0.0141, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, 0.0043, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0090, -0.0040, 0.0052, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0045, ..., 0.0302, -0.0037, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0041, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0020, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0045, -0.0222, 0.0262], + ..., + [-0.0147, 0.0051, 0.0260, ..., 0.0045, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0057, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0042, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0112], + [-0.0278, -0.0054, -0.0374, ..., 0.0276, -0.0046, 0.0151], + [ 0.0089, 0.0237, -0.0076, ..., 0.0527, -0.0078, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0075, -0.0095], + [-0.0068, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0039, -0.0128, ..., 0.0060, -0.0371, -0.0046], + [ 0.0071, -0.0127, 0.0045, ..., -0.0204, 0.0039, 0.0049], + [-0.0110, -0.0195, -0.0050, ..., -0.0542, -0.0092, -0.0063], + ..., + [-0.0205, -0.0084, 0.0039, ..., -0.0199, 0.0046, 0.0039], + [-0.0220, -0.0302, -0.0101, ..., 0.0105, 0.0369, 0.0041], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[-0.0042, 0.0347, 0.0092, ..., 0.0277, -0.0208, -0.0220], + [-0.0038, 0.0043, 0.0238, ..., -0.0145, -0.0320, -0.0046], + [-0.0193, 0.0364, 0.0039, ..., -0.0077, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0065, -0.0232, -0.0464], + [-0.0110, -0.0298, -0.0195, ..., 0.0112, 0.0312, 0.0154], + [-0.0072, -0.0119, -0.0195, ..., 0.0041, 0.0109, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0078, -0.0239, -0.0123, ..., -0.0062, 0.0243, -0.0396], + [-0.0100, 0.0154, 0.0026, ..., -0.0277, -0.0027, -0.0038], + ..., + [ 0.0117, -0.0052, 0.0515, ..., 0.0327, -0.0120, -0.0074], + [ 0.0039, 0.0170, 0.0055, ..., 0.0039, 0.0280, -0.0210], + [-0.0025, 0.0072, 0.0045, ..., 0.0155, 0.0042, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [ 0.0040, 0.0081, -0.0074, ..., -0.0588, 0.0187, -0.0078], + [-0.0102, 0.0199, -0.0078, ..., 0.0080, -0.0332, -0.0076], + ..., + [-0.0052, -0.0464, 0.0277, ..., -0.0143, 0.0198, 0.0054], + [ 0.0162, 0.0011, 0.0085, ..., -0.0104, -0.0066, 0.0297], + [-0.0269, 0.0131, 0.0164, ..., 0.0084, 0.0176, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0046, ..., -0.0505, -0.0332, 0.0154], + [ 0.0208, 0.0366, 0.0007, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, 0.0058, -0.0201, ..., 0.0330, 0.0309, -0.0064], + ..., + [ 0.0028, 0.0123, 0.0153, ..., 0.0039, 0.0034, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0042, 0.0175], + [-0.0244, -0.0062, -0.0300, ..., -0.0061, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, 0.0040, 0.0075, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0050, ..., 0.0310, 0.0088, -0.0039], + [-0.0078, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0087, -0.0109, -0.0237, ..., -0.0104, 0.0157, -0.0021], + [-0.0085, 0.0205, 0.0145, ..., 0.0262, -0.0195, 0.0095], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0070]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0029, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0029, 0.0034, -0.0093], + [ 0.0017, -0.0011, 0.0398, ..., -0.0304, -0.0030, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0039], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0048, 0.0291], + [ 0.0132, -0.0004, -0.0339, ..., 0.0065, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, 0.0024], + [-0.0166, -0.0043, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0064, 0.0020, ..., 0.0240, 0.0031, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0020], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0009, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0066, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0041], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, 0.0034, 0.0021], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, -0.0042, ..., -0.0164, -0.0040, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0027, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0053, -0.0231, -0.0255, ..., -0.0068, 0.0295, -0.0046], + [-0.0255, 0.0513, 0.0030, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0035, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0064, -0.0066, -0.0364, ..., -0.0175, 0.0306, -0.0077], + [-0.0051, -0.0195, -0.0027, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0078, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0078, -0.0193, -0.0140], + [-0.0166, -0.0078, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0134, -0.0044, 0.0137, ..., 0.0021, 0.0211, -0.0275], + [ 0.0051, -0.0076, -0.0234, ..., 0.0088, -0.0084, -0.0166], + [ 0.0225, -0.0251, 0.0039, ..., 0.0072, 0.0236, 0.0077], + ..., + [ 0.0184, 0.0145, -0.0041, ..., 0.0269, -0.0085, -0.0096], + [-0.0038, 0.0044, 0.0078, ..., -0.0260, 0.0267, 0.0150], + [ 0.0102, 0.0043, 0.0212, ..., 0.0195, -0.0157, 0.0124]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0085, -0.0270, 0.0251, ..., 0.0095, 0.0052, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.0.proj.1.weight': tensor([[ 0.0018, 0.0251, 0.0957, ..., -0.0535, 0.0286, -0.0262], + [-0.0195, 0.0112, -0.1108, ..., 0.0304, -0.0312, 0.0593], + [ 0.0017, 0.1309, -0.0078, ..., -0.0058, 0.0078, -0.0282], + ..., + [ 0.0039, -0.0038, -0.0136, ..., 0.0142, -0.0222, -0.0051], + [-0.0349, -0.0645, 0.0508, ..., 0.0364, 0.0630, 0.0566], + [ 0.0192, -0.0223, 0.0010, ..., 0.0115, 0.0510, -0.0084]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_proj.weight': tensor([[ 0.0182, 0.0081, 0.0123, ..., 0.0016, -0.0015, 0.0272], + [ 0.0183, -0.0024, 0.0045, ..., 0.0009, -0.0028, 0.0209], + [-0.0061, 0.0053, -0.0056, ..., -0.0055, -0.0039, -0.0149], + ..., + [-0.0089, -0.0040, -0.0085, ..., 0.0106, -0.0037, -0.0076], + [-0.0016, -0.0033, -0.0052, ..., -0.0043, -0.0039, -0.0052], + [ 0.0081, 0.0026, 0.0023, ..., 0.0082, -0.0038, 0.0005]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_norm.weight': tensor([ 1.0742e-01, -9.8633e-02, 1.1279e-01, 3.6914e-01, -4.8056e-07, + -1.9092e-07, 2.0752e-03, 7.8870e-13, 5.4956e-15, 4.9738e-13, + -2.0961e-13, 3.3307e-16, 1.1519e-15, -4.4631e-14, 1.0868e-10, + -4.3485e-12, -3.3617e-05, 1.4827e-06, -8.6212e-04, 5.6744e-05, + 1.1505e-10, -2.5466e-11, 1.9409e-02, -4.1962e-04, 3.0518e-02, + 6.6406e-01, 1.3867e-01, -2.4414e-03, 5.6641e-02, 1.0596e-01, + 1.3279e-10, 5.4845e-14, -2.2736e-03, 1.4663e-05, 2.9755e-03, + 1.0376e-03, 6.8212e-11, -2.9650e-10, 4.8637e-05, 1.5099e-14, + 9.0639e-17, -1.2434e-14, 6.4393e-15, 4.4250e-03, 3.2471e-02, + 2.2125e-03, -3.5400e-02, -4.6875e-02, 4.8242e-01, 2.6703e-03, + -6.8665e-03, 3.4027e-03, 3.9062e-03, 1.6838e-06, 2.1210e-03, + 1.2733e-10, -1.3131e-11, 3.3396e-13, -4.5103e-16, 1.0249e-19, + -1.2967e-13, -3.8858e-15, 2.6557e-10, 3.3168e-15, 3.3203e-01, + 3.4766e-01, 4.3750e-01, 1.1768e-01, -2.4959e-07, -1.0245e-07, + 6.9580e-03, 1.5135e-12, -2.9088e-14, -3.2152e-13, 2.8866e-14, + 3.3307e-16, -1.2663e-16, -2.4425e-14, 1.8758e-11, -6.3949e-13, + 7.3910e-05, 4.0770e-05, 2.8133e-05, 5.7220e-05, -5.3478e-10, + -1.2296e-09, -8.8120e-04, -1.3580e-03, 4.8633e-01, 9.0820e-02, + 6.0938e-01, 7.9590e-02, 4.5471e-03, -2.6550e-03, 5.6389e-11, + 1.0081e-13, 2.9373e-04, 1.6117e-04, -9.7752e-06, 4.9210e-04, + -5.7116e-10, 1.4734e-10, 3.7994e-03, -3.6915e-15, 7.5287e-16, + -4.1980e-16, 2.0400e-15, 2.5269e-02, -6.6223e-03, -1.1978e-03, + 4.5117e-01, 2.8516e-01, 1.0498e-01, -7.8125e-03, 7.8125e-03, + -2.9907e-03, -4.0894e-03, 2.6375e-06, 2.1057e-03, -1.4552e-10, + 3.4333e-11, -4.5119e-13, 1.4502e-15, -3.6321e-18, 4.6896e-13, + -9.6367e-14, 4.2917e-12, 3.7303e-14], dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.self_attn.k_proj.weight': tensor([[-0.0102, 0.0103, -0.0096, ..., -0.0046, -0.0105, 0.0022], + [-0.0124, 0.0089, 0.0037, ..., 0.0029, -0.0098, 0.0126], + [-0.0073, -0.0295, 0.0043, ..., -0.0023, 0.0043, -0.0208], + ..., + [-0.0018, 0.0002, -0.0042, ..., -0.0025, -0.0005, -0.0027], + [-0.0027, -0.0030, -0.0047, ..., 0.0026, 0.0041, 0.0038], + [-0.0037, -0.0039, -0.0026, ..., 0.0029, -0.0006, 0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.k_norm.weight': tensor([ 2.9102e-01, 2.0312e-01, -2.0508e-01, 3.4961e-01, 4.7497e-07, + -2.1793e-07, 4.3335e-03, 8.1357e-13, 1.1990e-14, 5.0449e-13, + 1.8741e-13, 2.9837e-16, 1.1519e-15, -4.4853e-14, -1.0823e-10, + -4.4054e-12, 7.4863e-05, -1.2945e-07, 1.0910e-03, 5.6744e-05, + 1.2187e-10, -2.9786e-11, 1.3367e-02, 7.2861e-04, 4.1260e-02, + 6.6406e-01, 1.9336e-01, 7.6660e-02, 4.1809e-03, 2.9206e-05, + -1.3733e-10, 6.8390e-14, 2.2888e-03, 4.6492e-05, 2.9907e-03, + 1.0376e-03, 7.9126e-11, 2.9286e-10, 6.8665e-05, 1.4322e-14, + 1.1241e-15, -6.6058e-15, -6.4670e-15, 2.1118e-02, 2.9297e-02, + 2.1820e-03, 1.0254e-01, 1.1963e-01, 5.0391e-01, 7.8125e-03, + -7.8125e-03, -3.9062e-03, 4.0894e-03, -3.2783e-06, -2.0142e-03, + -6.5938e-11, -1.4353e-12, 3.4639e-13, 2.3419e-16, -4.7095e-19, + 1.4655e-13, -7.8271e-15, -2.6739e-10, 3.7470e-15, 1.7285e-01, + 3.4180e-01, 4.3164e-01, -2.6172e-01, -2.3562e-07, -1.1583e-08, + 3.5553e-03, 1.4566e-12, -2.6978e-14, 3.3751e-13, 3.1752e-14, + -3.8858e-16, 1.0278e-16, 1.5543e-14, -1.9554e-11, -6.6436e-13, + 1.1349e-04, -4.7684e-05, 1.4603e-05, 5.8413e-05, 5.3842e-10, + -1.2151e-09, 2.1744e-04, -9.0408e-04, 4.8242e-01, -8.9355e-02, + 6.0156e-01, -3.7689e-03, 5.5664e-02, 1.0547e-01, 1.1880e-11, + 2.2093e-14, -3.5095e-04, 1.6117e-04, -1.9169e-04, -4.8256e-04, + 5.7844e-10, -2.5466e-10, 3.7689e-03, 3.7748e-15, 5.4678e-15, + -8.6736e-17, 2.0817e-15, 1.4282e-02, 1.7334e-02, 1.4725e-03, + 4.5312e-01, 2.6172e-01, 1.4453e-01, 4.6082e-03, 7.6294e-03, + 1.3199e-03, 3.0975e-03, 9.3877e-07, 2.2583e-03, 1.9281e-10, + 3.6607e-11, 4.1389e-13, -1.4919e-15, 3.5508e-18, -4.6896e-13, + -9.6811e-14, 1.1084e-12, 3.7303e-14], dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.self_attn.v_proj.weight': tensor([[-0.0022, 0.0098, 0.0055, ..., -0.0032, -0.0040, -0.0063], + [ 0.0002, 0.0038, -0.0006, ..., -0.0014, 0.0046, 0.0002], + [ 0.0142, 0.0043, 0.0038, ..., -0.0041, 0.0016, 0.0063], + ..., + [-0.0006, 0.0021, -0.0019, ..., 0.0025, -0.0025, -0.0029], + [ 0.0063, -0.0032, 0.0037, ..., 0.0059, -0.0058, 0.0040], + [-0.0090, 0.0049, -0.0024, ..., 0.0055, -0.0038, -0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.self_attn.output_proj.weight': tensor([[ 0.0145, -0.0216, -0.0175, ..., 0.0042, 0.0087, -0.0032], + [ 0.0022, -0.0068, 0.0124, ..., -0.0063, -0.0024, -0.0059], + [ 0.0040, 0.0022, 0.0052, ..., 0.0009, -0.0058, 0.0017], + ..., + [ 0.0055, -0.0001, 0.0048, ..., 0.0042, -0.0014, 0.0009], + [-0.0055, -0.0045, 0.0275, ..., 0.0036, -0.0033, 0.0161], + [-0.0084, 0.0095, -0.0052, ..., 0.0051, 0.0051, -0.0053]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_proj.weight': tensor([[-0.0034, -0.0093, -0.0017, ..., -0.0048, -0.0034, -0.0014], + [-0.0096, 0.0098, 0.0019, ..., 0.0072, 0.0025, -0.0030], + [-0.0023, -0.0084, 0.0025, ..., -0.0004, 0.0062, 0.0029], + ..., + [ 0.0030, 0.0071, 0.0009, ..., 0.0047, -0.0041, 0.0082], + [-0.0015, -0.0042, -0.0017, ..., -0.0048, 0.0043, -0.0067], + [-0.0060, 0.0023, 0.0009, ..., -0.0008, 0.0033, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_norm.weight': tensor([ 6.0120e-03, 8.8811e-06, 2.7740e-11, 1.4526e-02, -9.6436e-03, + 1.5442e-02, 2.2054e-06, 5.4550e-04, 3.4912e-02, 2.6978e-02, + 1.6880e-08, 2.8076e-02, 5.7812e-01, 9.3994e-03, 1.0437e-02, + 4.0894e-03, 4.6680e-01, 1.1086e-05, 1.1215e-03, 1.0803e-02, + 8.8501e-03, 7.3547e-03, 8.3496e-02, 1.8120e-05, 1.8433e-02, + 1.0681e-03, 7.9346e-03, -8.4400e-05, 1.2436e-03, 1.6708e-03, + 3.5156e-01, 9.0942e-03, 5.0306e-05, 4.6680e-01, 5.4297e-01, + 1.3828e-04, -5.1022e-05, 3.1494e-02, 1.5137e-01, 2.5635e-02, + 2.0996e-02, -4.6790e-06, 1.3642e-10, 4.1809e-03, 9.4223e-04, + 3.3203e-02, 1.2024e-02, 1.0605e-03, 4.0894e-03, 8.7891e-03, + 1.8539e-03, -7.4863e-05, -4.2915e-05, 1.5106e-03, -3.8300e-03, + 1.6689e-04, 1.9653e-02, 1.1396e-04, 1.7090e-02, 1.1475e-02, + 1.9932e-04, -7.8082e-06, 1.2329e-02, 7.2632e-03, 4.0293e-05, + 1.3550e-02, 9.9945e-04, 2.5000e-01, 8.8501e-04, 2.1851e-02, + 4.6875e-01, -1.2875e-05, 4.5410e-02, 8.4229e-03, 1.7695e-08, + 7.0190e-03, 1.6632e-03, 7.0801e-03, -1.3590e-05, 2.0752e-02, + 2.3438e-02, 4.2725e-03, 1.7188e-01, 7.6294e-03, 1.5869e-03, + 2.4567e-03, 1.3000e-02, -1.3530e-05, 1.0395e-04, 5.0781e-01, + 2.6550e-03, 8.9645e-05, 1.1520e-03, 3.8147e-06, 9.3384e-03, + 4.5117e-01, 4.3640e-03, 1.0538e-04, -3.7126e-13, 9.4223e-04, + 3.9062e-01, 1.3351e-05, -9.0332e-03, 1.7822e-02, 2.3071e-02, + 9.3460e-05, -1.5080e-05, 1.3769e-05, 1.6708e-03, 1.7944e-02, + 1.1597e-02, 8.5449e-04, 4.2419e-03, 1.2329e-02, -3.6316e-03, + 1.0395e-04, 1.1230e-02, -2.8534e-03, -1.1292e-03, 1.8311e-03, + 1.0300e-04, 6.3782e-03, 1.1368e-03, -2.2650e-05, 1.8978e-04, + 1.1414e-02, -7.1526e-06, 1.0133e-05], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_proj.weight': tensor([[ 1.5411e-03, 7.2861e-04, 1.9150e-03, ..., -6.7520e-04, + -2.4605e-04, 6.1035e-04], + [-3.6926e-03, -2.0599e-03, -3.9368e-03, ..., -2.6093e-03, + -3.9673e-03, -3.0060e-03], + [-1.7700e-03, -1.4877e-03, -2.4872e-03, ..., -2.8381e-03, + -5.6458e-03, -9.6130e-04], + ..., + [ 2.5635e-03, -2.4719e-03, 1.8768e-03, ..., 5.2643e-04, + 2.3041e-03, -5.7220e-04], + [-8.8882e-04, 5.3024e-04, -2.3193e-03, ..., -3.0212e-03, + 5.0354e-04, 2.3193e-03], + [ 2.9297e-03, 1.8997e-03, -3.2501e-03, ..., -4.2725e-03, + 6.5613e-04, 5.7697e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_norm.weight': tensor([ 5.9814e-03, -5.5730e-06, -2.7740e-11, 1.4343e-02, 9.7046e-03, + 1.5442e-02, 2.1756e-06, -5.4550e-04, 3.3936e-02, 2.7100e-02, + -5.9954e-09, 2.8320e-02, 5.8203e-01, 9.3384e-03, 9.8877e-03, + 4.0588e-03, 4.6680e-01, 1.1325e-05, 9.9182e-04, 1.0620e-02, + 8.7280e-03, 7.3853e-03, 8.3496e-02, 3.7670e-05, 1.8311e-02, + 1.0757e-03, 7.5073e-03, -8.3923e-05, 6.2561e-04, 1.6403e-03, + 3.5156e-01, 8.9111e-03, 5.0068e-05, 4.6680e-01, 5.4297e-01, + -2.0981e-04, 5.8413e-05, 3.1982e-02, 1.5137e-01, 2.5879e-02, + 2.0996e-02, 9.7275e-05, 1.9936e-09, 4.0283e-03, -2.2793e-04, + 3.2715e-02, 1.1902e-02, 1.0757e-03, 4.0283e-03, 8.7280e-03, + 1.8463e-03, -7.4863e-05, 4.2915e-05, 1.5030e-03, 3.9978e-03, + 2.1076e-04, 1.9531e-02, 1.3351e-04, 1.7334e-02, 1.1536e-02, + 1.9932e-04, -8.1062e-06, 1.2085e-02, 7.2632e-03, 4.0054e-05, + 1.3733e-02, 9.9182e-04, 2.5000e-01, 8.9645e-04, 2.1729e-02, + 4.6875e-01, 1.2696e-05, 4.3945e-02, 8.4229e-03, -1.9674e-08, + 7.1411e-03, 5.5313e-04, 7.0190e-03, -1.4782e-05, 2.0752e-02, + 2.3560e-02, 4.2419e-03, 1.7188e-01, 7.4463e-03, 1.6251e-03, + 2.4872e-03, 1.4038e-02, 1.3590e-05, 1.0490e-04, 5.0781e-01, + 2.6398e-03, -1.2398e-04, 7.5912e-04, 3.7253e-06, 9.3384e-03, + 4.5117e-01, -4.2419e-03, -9.1553e-05, 3.7126e-13, 9.1934e-04, + 3.9062e-01, 1.3351e-05, 1.0193e-02, 1.7822e-02, 2.3071e-02, + 9.2983e-05, 1.7524e-05, 1.3888e-05, 8.1253e-04, 1.7944e-02, + 1.1841e-02, 8.5449e-04, 4.2114e-03, 1.2268e-02, 3.6163e-03, + -1.6785e-04, 1.1292e-02, 2.8687e-03, 1.3123e-03, 1.8845e-03, + 9.7275e-05, 6.5918e-03, 1.1139e-03, -9.0599e-05, 1.0633e-04, + 1.1292e-02, -6.8247e-06, -3.7700e-06], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.v_proj.weight': tensor([[-9.8877e-03, 2.1820e-03, 1.2695e-02, ..., 2.8534e-03, + 2.6550e-03, -1.0193e-02], + [-2.2461e-02, 8.1635e-04, 7.7209e-03, ..., -3.2043e-03, + -6.1035e-03, 9.3937e-05], + [-1.2573e-02, 2.4567e-03, 2.9297e-03, ..., -2.2888e-03, + 1.3000e-02, -1.2268e-02], + ..., + [ 1.0925e-02, 1.8692e-03, 1.6861e-03, ..., 2.5940e-04, + 2.9755e-03, 1.8799e-02], + [-9.5215e-03, 2.5330e-03, 1.0315e-02, ..., 1.8463e-03, + -1.7700e-02, -3.7079e-03], + [-1.6479e-02, -2.8839e-03, 7.1716e-03, ..., -2.0599e-03, + -6.0425e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.output_proj.weight': tensor([[ 0.0064, -0.0018, 0.0056, ..., -0.0030, 0.0018, 0.0080], + [ 0.0080, 0.0066, -0.0012, ..., 0.0162, 0.0041, -0.0110], + [ 0.0052, -0.0030, -0.0008, ..., -0.0004, -0.0073, -0.0052], + ..., + [-0.0071, -0.0077, -0.0051, ..., -0.0079, -0.0047, -0.0085], + [ 0.0101, -0.0025, 0.0059, ..., -0.0142, -0.0183, 0.0055], + [ 0.0106, -0.0081, -0.0041, ..., -0.0070, -0.0037, -0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_img.weight': tensor([[ 0.0067, 0.0119, -0.0041, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0075, 0.0249, 0.0291], + [-0.0095, 0.0046, -0.0115, ..., -0.0342, -0.0327, 0.0061], + ..., + [ 0.0074, -0.0067, 0.0040, ..., 0.0082, 0.0286, -0.0165], + [ 0.0078, 0.0013, -0.0102, ..., -0.0100, 0.0142, -0.0309], + [ 0.0242, -0.0106, 0.0255, ..., -0.0227, -0.0043, -0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, -0.0010], + [ 0.0026, 0.0120, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0085, ..., -0.0057, -0.0046, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0039, 0.0054], + [-0.0160, 0.0055, -0.0036, ..., 0.0133, -0.0173, -0.0049], + [-0.0051, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img.weight': tensor([[-0.0510, -0.0021, -0.0364, ..., -0.0225, -0.0143, -0.0055], + [-0.0157, 0.0053, -0.0364, ..., 0.0099, 0.0117, -0.0036], + [ 0.0025, 0.0615, 0.0056, ..., -0.0205, -0.0306, -0.0114], + ..., + [ 0.0008, -0.0018, -0.0103, ..., -0.0211, -0.0012, 0.0288], + [ 0.0266, -0.0115, -0.0056, ..., -0.0420, -0.0282, -0.0008], + [ 0.0286, 0.0325, 0.0095, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.0.mlp.layer1.weight': tensor([[ 0.0082, 0.0159, 0.0067, ..., 0.0008, 0.0096, 0.0064], + [ 0.0050, -0.0152, -0.0035, ..., -0.0021, -0.0115, 0.0064], + [ 0.0062, -0.0057, -0.0039, ..., 0.0176, -0.0176, 0.0121], + ..., + [-0.0037, 0.0004, 0.0240, ..., 0.0054, 0.0075, 0.0029], + [-0.0030, 0.0186, 0.0068, ..., 0.0021, -0.0079, 0.0085], + [ 0.0050, 0.0027, -0.0013, ..., 0.0029, -0.0074, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.mlp.layer2.weight': tensor([[-0.0077, -0.0042, -0.0074, ..., -0.0038, -0.0261, -0.0078], + [ 0.0042, 0.0085, -0.0029, ..., -0.0162, -0.0216, 0.0073], + [-0.0062, 0.0092, 0.0090, ..., 0.0012, -0.0023, 0.0053], + ..., + [-0.0044, -0.0049, -0.0091, ..., -0.0069, -0.0026, -0.0082], + [ 0.0079, 0.0125, 0.0215, ..., 0.0008, 0.0056, 0.0114], + [ 0.0085, 0.0122, 0.0012, ..., 0.0050, -0.0021, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0610, -0.0593, -0.0579, ..., 0.0030, 0.0031, 0.0031], + [ 0.0591, 0.0583, 0.0562, ..., -0.0024, -0.0025, -0.0026], + [ 0.0583, 0.0576, 0.0564, ..., -0.0026, -0.0027, -0.0027], + ..., + [ 0.0630, 0.0615, 0.0598, ..., -0.0024, -0.0025, -0.0025], + [ 0.0598, 0.0583, 0.0566, ..., -0.0022, -0.0023, -0.0022], + [ 0.0598, 0.0583, 0.0571, ..., -0.0030, -0.0031, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0009, 0.0014, 0.0015, ..., 0.0050, 0.0036, 0.0003], + [-0.0025, -0.0015, -0.0016, ..., 0.0022, 0.0022, 0.0024], + [-0.0035, 0.0033, 0.0031, ..., 0.0059, 0.0060, 0.0034], + ..., + [-0.0432, 0.0500, 0.0493, ..., 0.0505, 0.0464, 0.0437], + [-0.0058, 0.0069, 0.0061, ..., 0.0053, 0.0044, 0.0057], + [ 0.0096, -0.0104, -0.0103, ..., -0.0099, -0.0096, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_cross_attn.1.weight': tensor([[-8.5449e-03, 4.4556e-03, 4.4861e-03, ..., -6.7520e-04, + -6.0654e-04, -7.7438e-04], + [-3.8818e-02, -3.7109e-02, -3.7109e-02, ..., 2.5177e-04, + 2.5940e-04, 2.6512e-04], + [ 4.1748e-02, 4.0283e-02, 4.0039e-02, ..., -7.0095e-05, + -5.1022e-05, -7.5340e-05], + ..., + [ 3.9062e-02, 3.8086e-02, 3.7109e-02, ..., -3.3569e-04, + -2.6703e-04, -3.0327e-04], + [ 4.9072e-02, 4.8096e-02, 4.6631e-02, ..., 5.0354e-04, + 6.1417e-04, 5.4550e-04], + [ 3.8818e-02, 3.7598e-02, 3.6621e-02, ..., -5.6624e-06, + 7.0572e-05, 2.1935e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0129, -0.0072, 0.0067, ..., 0.0072, 0.0072, 0.0068], + [ 0.0052, 0.0034, -0.0049, ..., -0.0034, -0.0034, -0.0048], + [-0.0037, -0.0036, 0.0035, ..., 0.0036, 0.0035, 0.0036], + ..., + [-0.0063, -0.0205, 0.0364, ..., 0.0203, 0.0432, 0.0320], + [ 0.0103, 0.0139, 0.0101, ..., -0.0131, 0.0050, 0.0099], + [-0.0137, -0.0046, -0.0113, ..., 0.0045, -0.0094, -0.0110]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_mlp.1.weight': tensor([[-0.0049, -0.0049, 0.0038, ..., 0.0004, 0.0007, 0.0006], + [ 0.0640, 0.0618, 0.0605, ..., -0.0024, -0.0027, -0.0027], + [ 0.0554, 0.0542, 0.0525, ..., -0.0037, -0.0040, -0.0040], + ..., + [ 0.0952, 0.0928, 0.0908, ..., -0.0008, -0.0009, -0.0010], + [-0.0038, -0.0038, -0.0038, ..., -0.0004, -0.0005, -0.0004], + [ 0.0747, 0.0723, 0.0708, ..., -0.0012, -0.0012, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.adaln_modulation_mlp.2.weight': tensor([[-0.0077, 0.0091, 0.0107, ..., 0.0072, 0.0082, 0.0060], + [-0.0007, 0.0023, 0.0033, ..., 0.0031, 0.0007, 0.0009], + [-0.0017, 0.0060, 0.0061, ..., 0.0044, 0.0015, 0.0057], + ..., + [ 0.0036, 0.0747, 0.1025, ..., 0.0635, -0.0044, 0.0496], + [ 0.0114, 0.0165, 0.0096, ..., 0.0280, -0.0114, 0.0233], + [-0.0089, 0.0033, -0.0003, ..., 0.0041, 0.0087, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.before_proj.weight': tensor([[ 0.0028, 0.0002, -0.0006, ..., 0.0062, 0.0011, -0.0013], + [-0.0094, -0.0055, -0.0041, ..., -0.0078, -0.0024, 0.0052], + [-0.0075, -0.0059, -0.0033, ..., -0.0015, -0.0036, -0.0027], + ..., + [-0.0011, 0.0017, -0.0047, ..., 0.0033, -0.0049, -0.0004], + [-0.0037, -0.0017, 0.0011, ..., -0.0020, -0.0052, -0.0042], + [ 0.0027, 0.0028, -0.0021, ..., 0.0045, 0.0005, -0.0034]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.before_proj.bias': tensor([-0.0199, 0.0027, 0.0018, ..., 0.0016, -0.0040, -0.0239], + dtype=torch.bfloat16), 'net.control_blocks_0.0.after_proj.weight': tensor([[-0.0188, -0.0083, 0.0017, ..., -0.0049, -0.0119, 0.0113], + [ 0.0189, -0.0134, 0.0063, ..., -0.0007, 0.0084, -0.0087], + [ 0.0009, -0.0009, -0.0043, ..., -0.0043, 0.0015, -0.0023], + ..., + [ 0.0029, -0.0039, 0.0083, ..., -0.0016, -0.0035, 0.0015], + [ 0.0276, -0.0150, 0.0077, ..., -0.0062, -0.0171, -0.0134], + [-0.0036, -0.0123, 0.0019, ..., 0.0026, -0.0048, 0.0112]], + dtype=torch.bfloat16), 'net.control_blocks_0.0.after_proj.bias': tensor([ 0.0075, -0.0165, -0.0026, ..., -0.0014, -0.0140, 0.0598], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_proj.weight': tensor([[-0.0220, 0.0138, -0.0125, ..., -0.0093, -0.0069, 0.0132], + [ 0.0049, 0.0032, -0.0035, ..., 0.0193, -0.0050, -0.0388], + [ 0.0012, 0.0103, 0.0013, ..., -0.0023, 0.0043, -0.0078], + ..., + [ 0.0009, 0.0050, -0.0006, ..., 0.0024, -0.0010, -0.0002], + [ 0.0021, -0.0055, -0.0013, ..., 0.0013, -0.0062, 0.0017], + [ 0.0025, -0.0006, 0.0052, ..., 0.0031, -0.0019, -0.0037]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_norm.weight': tensor([ 3.9062e-01, 1.2578e+00, 5.6396e-02, 1.1953e+00, 8.9844e-01, + 4.6945e-34, -1.9184e-34, -1.5573e-34, -6.8310e-34, 3.3087e-22, + 8.9676e-34, -4.9653e-34, -9.3287e-35, -3.7014e-34, -7.2523e-34, + -1.0653e-33, 1.4551e-01, 3.8338e-04, -2.1696e-05, 2.2339e-02, + 2.4319e-04, -5.7697e-05, 4.6289e-01, 8.8672e-01, 1.3516e+00, + 4.4531e-01, -8.6426e-02, 6.0547e-01, 8.4229e-03, -2.5022e-30, + 1.9259e-34, 1.6852e-33, -2.3127e-05, 8.9264e-04, 8.7738e-05, + -9.7656e-03, 3.1281e-04, -1.0909e-34, -4.9277e-35, 1.1977e-33, + -1.3542e-33, 2.1507e-22, 6.3496e-34, 1.0859e+00, 3.2422e-01, + 2.1973e-01, 3.7500e-01, 1.3438e+00, 2.0630e-02, -4.4556e-03, + -6.2256e-03, 2.3926e-01, 1.7395e-03, -4.6042e-34, -3.7917e-34, + -3.0333e-33, 7.4158e-03, -6.0100e-25, 1.9019e-33, 2.0613e-34, + 3.8556e-35, 4.0813e-35, -1.6325e-34, -2.7836e-35, 5.0391e-01, + 4.1406e-01, 6.3281e-01, 4.9805e-01, 1.6797e-01, 5.2662e-34, + -1.2883e-35, 4.4838e-34, 7.4028e-34, -3.2467e-23, 2.1667e-34, + -1.5528e-33, 2.7836e-35, -1.0833e-34, 9.3287e-34, 8.1852e-34, + 4.6387e-03, 8.9264e-04, -4.4346e-05, 4.4250e-04, 2.3079e-04, + 3.5889e-02, 1.1953e+00, 2.5391e-01, 4.5117e-01, 1.3359e+00, + 2.7734e-01, 2.4414e-01, 1.9455e-03, -2.0338e-30, 2.1215e-34, + -1.3120e-33, -3.0279e-05, 3.4714e-04, -1.6212e-05, 1.3550e-02, + 3.8719e-04, -1.1887e-34, -4.1001e-35, -1.6777e-34, -1.1857e-33, + -4.1359e-24, 4.2130e-33, 2.8711e-01, 9.4141e-01, 1.2266e+00, + 1.4844e+00, 1.9336e-01, 8.3594e-01, -3.9062e-03, -8.2397e-04, + -4.7913e-03, 2.5558e-04, 1.3181e-33, -3.4005e-34, 2.1787e-33, + 6.7520e-04, -7.9616e-24, 2.3232e-33, 1.7379e-34, -1.0909e-35, + 1.1134e-34, -6.9213e-35, 6.3947e-35], dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.self_attn.k_proj.weight': tensor([[ 0.0016, -0.0195, 0.0078, ..., 0.0172, -0.0051, -0.0264], + [ 0.0066, -0.0129, -0.0035, ..., -0.0199, 0.0034, -0.0095], + [ 0.0092, 0.0063, -0.0018, ..., 0.0054, -0.0044, 0.0065], + ..., + [-0.0085, 0.0008, -0.0060, ..., 0.0082, 0.0109, -0.0054], + [-0.0095, -0.0052, 0.0011, ..., -0.0036, 0.0062, -0.0051], + [-0.0058, -0.0028, -0.0025, ..., -0.0032, 0.0117, -0.0054]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.k_norm.weight': tensor([ 5.0391e-01, 1.2812e+00, 2.3438e-01, 1.2109e+00, 8.9844e-01, + 4.5139e-34, 1.3918e-34, 1.1586e-34, 1.5167e-33, 3.3087e-22, + -2.8137e-34, -2.0162e-34, 4.4387e-35, 1.9410e-34, -6.7708e-34, + 6.0938e-35, 1.4551e-01, 3.6812e-04, 1.2875e-04, 5.5420e-02, + 2.4319e-04, 1.2493e-04, 6.8750e-01, 8.9062e-01, 1.3516e+00, + 4.5117e-01, 2.1289e-01, 6.4453e-01, -8.8501e-03, 1.6147e-30, + 1.0382e-34, -7.8993e-35, 1.4782e-04, 8.6975e-04, 1.5545e-04, + 4.5410e-02, 3.0708e-04, -5.0781e-35, -2.4262e-35, 4.1829e-34, + -3.9120e-34, -8.5200e-23, 4.7245e-34, 1.0781e+00, 3.8672e-01, + -6.3477e-02, -3.8672e-01, 1.3359e+00, 1.5527e-01, 2.4567e-03, + -6.1646e-03, 2.3340e-01, 1.7395e-03, -2.5278e-34, 1.2940e-34, + 1.1857e-33, 7.1411e-03, -6.0746e-25, -3.3102e-34, 1.3015e-34, + -1.2413e-35, -6.1314e-35, 1.6852e-34, -2.2569e-35, 5.1562e-01, + 5.5078e-01, 6.2891e-01, 4.3359e-01, -2.4414e-01, -6.9514e-34, + 1.1360e-34, 2.0162e-34, -3.6352e-33, 1.5406e-23, -2.9190e-34, + -4.9051e-34, 1.9184e-35, 8.0874e-35, -2.1366e-34, 5.5972e-34, + 5.1575e-03, 8.8501e-04, 1.3161e-04, 4.2725e-04, 2.3460e-04, + 3.5889e-02, 1.1875e+00, 3.1836e-01, 4.7070e-01, 1.3438e+00, + 2.7539e-01, 3.0273e-01, 8.3923e-04, -1.8612e-30, 1.5799e-34, + -3.1296e-33, 1.0633e-04, 4.0627e-04, 4.3869e-05, 4.4678e-02, + 3.8528e-04, 3.6864e-35, -4.5515e-35, -1.3993e-34, -1.0472e-33, + 3.0632e-24, 2.0222e-33, 4.6094e-01, 9.4531e-01, 1.2344e+00, + 1.4844e+00, 2.3438e-01, 8.4375e-01, -6.0425e-03, -7.6294e-05, + -2.0703e-01, 3.1090e-04, -6.5000e-34, -3.7014e-34, -1.4264e-33, + 1.1587e-04, 3.5931e-24, 4.6945e-34, -1.0006e-34, -3.7616e-35, + -1.0382e-34, 3.2162e-35, -4.7396e-35], dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.self_attn.v_proj.weight': tensor([[ 0.0028, -0.0034, 0.0063, ..., -0.0203, -0.0049, 0.0070], + [ 0.0029, -0.0100, 0.0068, ..., 0.0069, -0.0041, -0.0063], + [ 0.0052, -0.0027, -0.0053, ..., -0.0231, -0.0046, -0.0111], + ..., + [ 0.0065, 0.0164, 0.0194, ..., -0.0010, 0.0004, -0.0175], + [-0.0047, -0.0121, 0.0077, ..., 0.0072, -0.0036, 0.0110], + [ 0.0112, -0.0040, -0.0160, ..., 0.0114, 0.0090, 0.0106]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.self_attn.output_proj.weight': tensor([[-0.0054, -0.0135, -0.0113, ..., 0.0189, -0.0115, 0.0073], + [ 0.0178, -0.0056, -0.0044, ..., -0.0254, -0.0120, -0.0076], + [-0.0027, 0.0036, 0.0038, ..., -0.0048, 0.0053, 0.0042], + ..., + [ 0.0027, -0.0137, -0.0098, ..., -0.0210, -0.0160, -0.0188], + [ 0.0042, -0.0032, -0.0053, ..., -0.0211, 0.0063, -0.0063], + [ 0.0143, 0.0056, -0.0114, ..., -0.0033, 0.0105, -0.0081]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_proj.weight': tensor([[-0.0164, 0.0239, 0.0057, ..., 0.0036, -0.0100, 0.0140], + [ 0.0244, -0.0101, -0.0052, ..., -0.0085, 0.0110, -0.0208], + [ 0.0104, 0.0075, -0.0023, ..., -0.0057, 0.0073, -0.0005], + ..., + [ 0.0092, -0.0036, 0.0033, ..., 0.0065, -0.0099, 0.0070], + [-0.0061, -0.0043, -0.0024, ..., -0.0044, -0.0095, 0.0100], + [ 0.0219, 0.0141, -0.0095, ..., -0.0045, 0.0086, -0.0172]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_norm.weight': tensor([ 2.3682e-02, 4.8047e-01, 1.7822e-02, 8.7891e-03, 6.7749e-03, + 2.0606e-12, 9.5367e-06, 5.5469e-01, -1.8332e-12, 4.4727e-01, + 1.8799e-02, 1.9287e-02, 7.0496e-03, 2.2095e-02, 1.7944e-02, + 1.7456e-02, 6.0425e-03, 1.8215e-04, -3.6001e-05, 2.9297e-02, + 2.6562e-01, 2.2461e-02, 6.5231e-04, 1.8433e-02, 1.0757e-03, + 2.6131e-04, -1.7548e-04, 3.0884e-02, 6.5327e-05, 4.2969e-01, + -1.1969e-04, -4.2915e-05, 9.6436e-03, 1.4114e-04, 7.6904e-03, + 1.0376e-02, 2.0146e-05, 4.5703e-01, 1.5259e-02, 2.7771e-03, + 3.6377e-02, 1.8921e-02, 5.9204e-03, 2.8133e-05, 5.4321e-03, + 7.6294e-03, 1.8616e-03, 1.7578e-02, 2.8076e-02, 1.4893e-02, + 4.9438e-03, 4.4336e-01, 1.0071e-02, 1.5869e-02, 3.1494e-02, + 1.8692e-04, 6.8359e-02, 1.6479e-02, 1.5869e-02, 2.5391e-02, + 4.0770e-05, 1.1475e-02, 5.7373e-03, -4.8876e-05, 1.2756e-02, + 2.5482e-03, -1.8597e-05, -2.7776e-05, 8.9722e-03, 4.7852e-02, + -5.2387e-08, 6.1951e-03, 4.6143e-02, 3.2471e-02, 4.3640e-03, + 8.3008e-03, 2.9755e-03, 1.7578e-02, 2.1973e-02, 1.4648e-02, + -2.2793e-04, 1.1963e-02, 3.2471e-02, -2.8253e-05, 2.9419e-02, + 1.2360e-03, 4.3030e-03, 3.0518e-02, 1.1658e-02, 5.5847e-03, + 1.3306e-02, 5.9891e-04, 1.1292e-02, 2.1729e-02, 1.3306e-02, + -1.0395e-04, 1.9287e-02, 3.3203e-02, 1.5991e-02, 4.1748e-02, + 2.5513e-02, 2.3438e-02, 1.6357e-02, 1.0443e-04, -2.0373e-10, + 1.3733e-02, 4.6158e-04, 2.2461e-02, 1.6479e-02, 1.9531e-02, + 1.3184e-02, 5.8594e-03, 2.1118e-02, 5.0781e-01, 1.1063e-03, + 1.1353e-02, 7.8735e-03, 1.2512e-02, 5.1575e-03, 5.2002e-02, + 2.3499e-03, 2.3651e-03, -1.4210e-04, 1.2451e-02, 6.4087e-03, + 1.3367e-02, 2.0264e-02, 2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_proj.weight': tensor([[-0.0010, -0.0033, 0.0038, ..., -0.0006, 0.0034, 0.0026], + [ 0.0116, -0.0035, -0.0040, ..., 0.0050, 0.0197, 0.0012], + [ 0.0023, 0.0036, 0.0016, ..., -0.0006, -0.0019, 0.0029], + ..., + [-0.0017, -0.0049, -0.0036, ..., -0.0004, -0.0015, 0.0031], + [-0.0016, -0.0053, -0.0125, ..., 0.0029, 0.0045, 0.0010], + [ 0.0078, -0.0004, -0.0020, ..., -0.0027, 0.0015, -0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_norm.weight': tensor([ 2.4658e-02, 4.8047e-01, 1.7578e-02, 8.8501e-03, 6.7749e-03, + 2.1885e-12, 9.4771e-06, 5.5469e-01, 1.8474e-12, 4.4727e-01, + 1.9409e-02, 1.9165e-02, 7.0190e-03, 2.1973e-02, 1.7822e-02, + 1.7334e-02, 6.2866e-03, 1.7929e-04, -3.7432e-05, 2.9175e-02, + 2.6562e-01, 2.1973e-02, 6.5613e-04, 1.8555e-02, 1.0986e-03, + 2.6321e-04, -1.7357e-04, 3.0640e-02, 6.5327e-05, 4.2969e-01, + -1.1921e-04, -4.4823e-05, 9.4604e-03, 1.4019e-04, 7.6904e-03, + 1.0315e-02, 2.0504e-05, 4.5703e-01, 1.5259e-02, 2.6550e-03, + 3.5400e-02, 1.8921e-02, 5.5847e-03, 2.5511e-05, 5.3406e-03, + 7.6904e-03, 1.8768e-03, 1.7334e-02, 2.8198e-02, 1.5015e-02, + 4.8828e-03, 4.4336e-01, 1.0132e-02, 1.5747e-02, 3.1250e-02, + 1.8597e-04, 6.8848e-02, 1.6235e-02, 1.5869e-02, 2.5513e-02, + 4.1485e-05, 1.1536e-02, 5.7373e-03, -5.0783e-05, 1.2756e-02, + 2.4872e-03, -1.9431e-05, -2.7418e-05, 9.0942e-03, 4.7852e-02, + -1.1316e-07, 6.3477e-03, 4.6631e-02, 3.2471e-02, 4.4250e-03, + 8.6670e-03, 2.9297e-03, 1.7090e-02, 2.1973e-02, 1.4771e-02, + -2.2316e-04, 1.1780e-02, 3.1982e-02, -2.7776e-05, 2.9663e-02, + 1.0910e-03, 4.2725e-03, 3.0273e-02, 1.1963e-02, 5.7373e-03, + 1.3794e-02, 6.2561e-04, 1.1292e-02, 2.1606e-02, 1.3123e-02, + -1.0586e-04, 1.8677e-02, 3.2959e-02, 1.5869e-02, 4.1016e-02, + 2.6367e-02, 2.3560e-02, 1.6602e-02, 1.0347e-04, 2.0373e-10, + 1.3672e-02, 4.5013e-04, 2.2217e-02, 1.6602e-02, 1.9287e-02, + 1.3184e-02, 5.5847e-03, 2.0996e-02, 5.0781e-01, 1.0986e-03, + 1.1292e-02, 7.8735e-03, 1.2329e-02, 5.3711e-03, 5.2002e-02, + 2.3956e-03, 2.3956e-03, -1.4210e-04, 1.2634e-02, 6.3477e-03, + 1.3245e-02, 2.0264e-02, 2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.v_proj.weight': tensor([[-2.2583e-02, 2.2736e-03, 2.2430e-03, ..., -8.6784e-05, + -1.5442e-02, -2.7832e-02], + [-4.1199e-03, 1.0803e-02, 1.0376e-02, ..., 7.7820e-03, + 6.6833e-03, 7.7209e-03], + [-3.9307e-02, -5.0049e-03, -7.4158e-03, ..., -9.3994e-03, + -2.7710e-02, -1.2451e-02], + ..., + [-4.1748e-02, 7.0190e-04, -1.5625e-02, ..., 1.8158e-03, + 2.1729e-02, 9.3994e-03], + [ 6.5308e-03, 6.1646e-03, -4.4556e-03, ..., 3.6469e-03, + 8.6670e-03, 8.1787e-03], + [ 1.6235e-02, 4.6692e-03, -3.8757e-03, ..., 2.0294e-03, + 1.1353e-02, -4.6015e-05]], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.output_proj.weight': tensor([[ 0.0071, 0.0040, 0.0148, ..., 0.0079, 0.0043, -0.0017], + [-0.0080, -0.0045, -0.0041, ..., -0.0087, 0.0038, 0.0112], + [-0.0005, 0.0038, -0.0038, ..., 0.0015, -0.0027, -0.0022], + ..., + [ 0.0054, 0.0052, 0.0035, ..., -0.0064, -0.0051, 0.0078], + [ 0.0128, 0.0302, -0.0167, ..., -0.0222, 0.0045, 0.0425], + [-0.0012, -0.0087, -0.0229, ..., -0.0104, 0.0020, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_img.weight': tensor([[ 0.0079, 0.0020, 0.0430, ..., -0.0327, 0.0054, 0.0148], + [-0.0278, 0.0280, 0.0249, ..., -0.0068, -0.0378, 0.0271], + [ 0.0139, -0.0116, 0.0312, ..., -0.0075, 0.0253, -0.0110], + ..., + [ 0.0245, -0.0150, 0.0033, ..., -0.0452, 0.0216, -0.0209], + [-0.0081, 0.0073, -0.0471, ..., -0.0109, 0.0157, -0.0088], + [ 0.0078, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0038, -0.0123, -0.0115], + [ 0.0135, 0.0061, -0.0188, ..., 0.0157, 0.0082, -0.0042], + [-0.0306, -0.0028, -0.0272, ..., 0.0101, 0.0337, 0.0039], + ..., + [-0.0040, -0.0040, 0.0047, ..., 0.0209, -0.0103, 0.0073], + [ 0.0193, 0.0210, 0.0123, ..., -0.0145, 0.0117, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0110, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img.weight': tensor([[ 0.0058, -0.0056, 0.0173, ..., 0.0083, 0.0356, 0.0217], + [-0.0498, -0.0051, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0108, -0.0044, -0.0139, ..., 0.0039, 0.0269, 0.0077], + ..., + [-0.0043, 0.0125, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0077, -0.0162, ..., -0.0054, -0.0075, -0.0121], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0138, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.1.mlp.layer1.weight': tensor([[-0.0176, 0.0156, -0.0038, ..., -0.0092, -0.0087, -0.0221], + [-0.0026, -0.0106, 0.0096, ..., -0.0039, -0.0060, -0.0166], + [-0.0065, 0.0032, 0.0022, ..., -0.0078, -0.0032, -0.0050], + ..., + [ 0.0251, 0.0005, -0.0049, ..., 0.0010, 0.0085, 0.0071], + [-0.0251, 0.0047, -0.0145, ..., 0.0160, -0.0053, 0.0062], + [ 0.0092, -0.0067, -0.0008, ..., -0.0135, 0.0120, -0.0248]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.mlp.layer2.weight': tensor([[-0.0166, -0.0135, 0.0115, ..., -0.0025, 0.0060, -0.0061], + [-0.0049, 0.0071, 0.0084, ..., -0.0048, 0.0110, 0.0041], + [-0.0077, -0.0070, 0.0029, ..., -0.0098, 0.0082, -0.0077], + ..., + [-0.0027, 0.0090, 0.0045, ..., 0.0143, -0.0068, 0.0093], + [ 0.0049, 0.0103, 0.0046, ..., -0.0059, -0.0005, 0.0159], + [ 0.0105, -0.0038, -0.0099, ..., 0.0050, -0.0029, -0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0447, -0.0435, -0.0430, ..., 0.0050, 0.0053, 0.0053], + [ 0.0408, 0.0396, 0.0381, ..., -0.0037, -0.0036, -0.0037], + [-0.0454, -0.0437, -0.0432, ..., 0.0051, 0.0052, 0.0051], + ..., + [-0.0498, -0.0488, -0.0476, ..., 0.0035, 0.0036, 0.0037], + [-0.0439, -0.0432, -0.0425, ..., 0.0037, 0.0040, 0.0040], + [-0.0310, -0.0295, -0.0284, ..., 0.0011, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0058, 0.0006, 0.0046, ..., 0.0025, 0.0046, -0.0006], + [ 0.0018, -0.0023, -0.0004, ..., 0.0001, 0.0035, 0.0018], + [-0.0065, -0.0017, -0.0055, ..., -0.0077, -0.0070, 0.0017], + ..., + [-0.0293, 0.0015, -0.0344, ..., -0.0378, -0.0361, -0.0017], + [-0.0250, -0.0064, -0.0226, ..., -0.0249, -0.0222, 0.0072], + [ 0.0065, -0.0074, 0.0048, ..., 0.0065, -0.0005, 0.0062]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_cross_attn.1.weight': tensor([[ 6.6406e-02, 6.4941e-02, 6.3965e-02, ..., 4.8161e-05, + 1.6093e-05, 5.9843e-05], + [ 4.4556e-03, 4.4250e-03, 4.4250e-03, ..., -3.4332e-05, + -1.1063e-04, -2.2602e-04], + [-2.2827e-02, -2.1973e-02, -2.0996e-02, ..., -6.8283e-04, + -6.8283e-04, -6.9427e-04], + ..., + [-2.7344e-02, -2.6978e-02, -2.6245e-02, ..., -4.1580e-04, + -3.7575e-04, -3.3760e-04], + [-2.5391e-02, -2.4902e-02, -2.4170e-02, ..., -2.0599e-04, + -1.6022e-04, -2.1267e-04], + [ 5.8838e-02, 5.7617e-02, 5.6885e-02, ..., -1.4591e-04, + -2.6822e-05, -1.7166e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0094, -0.0076, -0.0112, ..., -0.0010, 0.0064, -0.0079], + [ 0.0042, 0.0039, 0.0047, ..., -0.0049, -0.0038, 0.0056], + [ 0.0130, -0.0038, -0.0038, ..., 0.0047, -0.0058, 0.0097], + ..., + [ 0.0679, 0.0107, -0.0051, ..., 0.0074, -0.0149, 0.0483], + [ 0.0288, -0.0056, -0.0077, ..., -0.0071, -0.0209, 0.0310], + [-0.0078, 0.0099, -0.0190, ..., 0.0144, 0.0126, -0.0093]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_mlp.1.weight': tensor([[-0.0043, -0.0045, -0.0044, ..., 0.0003, 0.0003, 0.0003], + [ 0.0659, 0.0640, 0.0635, ..., -0.0038, -0.0039, -0.0038], + [ 0.0415, 0.0408, 0.0403, ..., -0.0039, -0.0039, -0.0039], + ..., + [ 0.0320, 0.0309, 0.0306, ..., -0.0021, -0.0022, -0.0021], + [-0.0124, -0.0114, -0.0105, ..., 0.0014, 0.0014, 0.0013], + [ 0.0254, 0.0253, 0.0253, ..., -0.0025, -0.0026, -0.0026]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.adaln_modulation_mlp.2.weight': tensor([[-0.0028, 0.0018, 0.0028, ..., 0.0026, 0.0021, 0.0024], + [-0.0002, 0.0020, 0.0021, ..., 0.0019, -0.0001, 0.0010], + [ 0.0019, 0.0001, -0.0037, ..., -0.0031, -0.0019, -0.0029], + ..., + [-0.0029, 0.0013, -0.0189, ..., -0.0081, 0.0028, -0.0057], + [ 0.0151, -0.0075, -0.0244, ..., -0.0165, -0.0144, -0.0148], + [ 0.0172, -0.0195, -0.0133, ..., -0.0166, -0.0178, -0.0162]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.after_proj.weight': tensor([[ 0.0104, -0.0075, 0.0262, ..., -0.0078, 0.0193, -0.0015], + [-0.0023, -0.0077, 0.0088, ..., 0.0153, 0.0086, -0.0063], + [ 0.0081, 0.0028, -0.0168, ..., 0.0070, 0.0148, -0.0087], + ..., + [-0.0050, 0.0074, 0.0249, ..., -0.0206, -0.0128, -0.0078], + [-0.0036, 0.0177, 0.0115, ..., -0.0117, 0.0003, 0.0161], + [-0.0123, -0.0178, -0.0145, ..., 0.0075, 0.0110, -0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.1.after_proj.bias': tensor([-4.2419e-03, 5.1880e-03, 4.8523e-03, ..., 6.7749e-03, + 7.0572e-05, -1.3885e-03], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_proj.weight': tensor([[ 0.0249, -0.0265, -0.0237, ..., 0.0074, -0.0041, 0.0330], + [ 0.0048, 0.0044, -0.0016, ..., -0.0142, 0.0068, -0.0049], + [ 0.0253, 0.0031, -0.0038, ..., 0.0045, -0.0076, 0.0183], + ..., + [-0.0087, -0.0070, 0.0021, ..., -0.0044, -0.0036, -0.0004], + [ 0.0315, 0.0259, -0.0068, ..., -0.0079, -0.0048, 0.0042], + [-0.0124, -0.0121, 0.0179, ..., -0.0028, 0.0063, -0.0244]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_norm.weight': tensor([ 1.3750e+00, 5.2344e-01, 6.6406e-01, 4.7852e-01, 1.3359e+00, + 1.4893e-02, 9.6436e-03, 1.9789e-05, 2.1057e-03, 1.5137e-02, + 3.3677e-06, -7.6252e-09, 1.1406e+00, -2.5988e-05, -3.5508e-18, + -4.7119e-02, 1.1597e-02, 2.6489e-02, 2.6703e-04, 1.0010e-02, + 1.5527e-01, -3.7909e-05, 9.6875e-01, 1.5938e+00, 7.1094e-01, + 1.1016e+00, 7.1094e-01, 1.9141e-01, 2.3682e-02, -1.4099e-02, + 7.1106e-03, 3.1090e-04, 7.5989e-03, -3.0212e-03, -1.0071e-03, + 2.9663e-02, 1.3965e-01, -2.5272e-05, 5.9814e-03, 2.1973e-03, + -7.0930e-06, -6.2883e-06, -1.7956e-06, 1.5625e+00, 6.4844e-01, + 1.5000e+00, 1.0625e+00, 4.8633e-01, -3.2227e-02, -3.9673e-03, + -9.9182e-04, 2.7222e-02, 1.3184e-01, -3.9101e-05, 6.4373e-05, + -4.3213e-06, -2.4033e-04, -5.0659e-03, -5.5176e-02, -3.1832e-10, + 3.2471e-02, -1.6809e-05, -2.2461e-02, -2.8320e-02, 8.9844e-01, + 1.2422e+00, 1.3438e+00, 1.3828e+00, 6.1328e-01, -2.0508e-02, + 9.5215e-03, 8.8215e-05, 1.2024e-02, 1.2283e-03, 9.0525e-07, + 1.7695e-07, 6.6895e-02, 9.3460e-05, 9.1903e-20, 1.0000e+00, + 1.1797e+00, 2.8564e-02, 2.2507e-04, 3.1494e-02, 1.1475e-01, + 1.1475e-02, 1.6719e+00, 9.8438e-01, 1.4609e+00, 1.5391e+00, + 2.9102e-01, 6.1328e-01, 5.2490e-02, -8.4229e-03, -1.1719e-02, + 1.7357e-04, -3.0396e-02, -1.8677e-02, -3.6316e-03, 4.1260e-02, + 1.3281e-01, -4.4584e-05, 1.1875e+00, -3.6478e-05, -9.1195e-06, + -9.1553e-05, -5.9372e-08, 1.1875e+00, 1.6328e+00, 1.5469e+00, + 1.5703e+00, 8.1250e-01, 1.9775e-02, -2.6001e-02, 3.2959e-03, + 3.4424e-02, 1.3770e-01, 1.4544e-05, 4.1723e-05, -3.1441e-06, + -4.1723e-06, -6.1512e-05, 1.2266e+00, -8.3674e-11, 3.0640e-02, + -5.6028e-06, 9.6484e-01, 1.2188e+00], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.self_attn.k_proj.weight': tensor([[-0.0034, 0.0308, 0.0071, ..., 0.0121, 0.0184, 0.0039], + [ 0.0065, 0.0037, 0.0056, ..., -0.0039, 0.0052, 0.0047], + [-0.0232, 0.0197, 0.0015, ..., 0.0042, 0.0145, 0.0150], + ..., + [ 0.0044, -0.0096, -0.0056, ..., 0.0079, 0.0087, 0.0111], + [ 0.0140, 0.0077, 0.0138, ..., -0.0079, -0.0084, 0.0176], + [ 0.0022, -0.0189, 0.0095, ..., -0.0082, 0.0088, 0.0299]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.k_norm.weight': tensor([ 1.4141e+00, 7.5781e-01, 4.7070e-01, 4.7461e-01, 1.3438e+00, + 1.5991e-02, 1.3977e-02, 8.8692e-05, 7.1716e-03, 1.5076e-02, + 3.2634e-06, -4.3656e-09, 1.1406e+00, 4.3869e-05, -3.5779e-18, + -3.7537e-03, -1.1206e-04, 9.2773e-02, 3.2196e-03, -3.2196e-03, + 4.0283e-03, -3.9339e-05, 1.0625e+00, 1.4609e+00, 4.9609e-01, + 1.2266e+00, 7.8125e-01, 2.9102e-01, 1.2451e-02, 5.8289e-03, + 1.0254e-02, 3.3379e-04, 8.3008e-03, 4.3213e-02, 1.8539e-03, + 9.0332e-03, 2.7710e-02, 1.7643e-05, 2.1729e-02, 2.1973e-03, + -8.8811e-06, -2.3395e-06, 1.6317e-06, 1.7188e+00, 1.1875e+00, + 1.5156e+00, 1.2734e+00, 4.8828e-01, -1.6357e-02, 6.4453e-02, + 3.1281e-03, -9.9487e-03, 4.7913e-03, -3.2902e-05, -1.7047e-05, + 3.9935e-06, -2.5177e-04, -5.0659e-03, 3.0640e-02, 2.7831e-10, + -3.9673e-03, -1.6212e-05, -3.0029e-02, -2.9907e-02, 6.8359e-01, + 1.1406e+00, 1.3750e+00, 1.3594e+00, 4.8242e-01, -1.9165e-02, + 6.1035e-03, -4.5300e-06, 3.5553e-03, -9.6893e-04, 1.1921e-06, + 1.8068e-07, 5.3406e-03, 5.8889e-05, 6.7424e-19, 1.0000e+00, + 1.1797e+00, 8.1055e-02, -4.8828e-04, -3.8300e-03, -3.6316e-03, + 1.1597e-02, 1.5859e+00, 1.2266e+00, 1.5703e+00, 1.5703e+00, + 8.7402e-02, 6.0547e-01, 2.3193e-02, 1.3977e-02, -7.9346e-03, + 1.0395e-04, 3.1738e-02, 6.0303e-02, 7.0801e-03, -1.0254e-02, + -2.5879e-02, -5.1022e-05, 1.1875e+00, 3.6478e-05, -2.4199e-05, + -9.1076e-05, -6.9384e-08, 1.0469e+00, 1.4766e+00, 1.4609e+00, + 1.6797e+00, 8.2031e-01, -2.5024e-02, 9.8633e-02, 3.6163e-03, + 1.9165e-02, -1.6602e-02, -2.7657e-05, 3.2425e-05, -2.9951e-06, + -2.0266e-06, 7.4387e-05, 1.2266e+00, -2.8081e-11, -1.0742e-02, + 5.3346e-06, 9.6484e-01, 1.2188e+00], dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.self_attn.v_proj.weight': tensor([[-0.0051, 0.0036, -0.0222, ..., -0.0315, 0.0049, 0.0051], + [ 0.0030, 0.0017, -0.0138, ..., 0.0074, 0.0099, 0.0056], + [-0.0173, -0.0101, 0.0111, ..., 0.0151, -0.0081, -0.0086], + ..., + [ 0.0072, 0.0109, -0.0194, ..., -0.0356, 0.0075, 0.0027], + [ 0.0049, 0.0150, -0.0041, ..., 0.0060, 0.0060, -0.0049], + [ 0.0119, -0.0081, -0.0173, ..., 0.0171, -0.0041, 0.0049]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.self_attn.output_proj.weight': tensor([[ 0.0085, 0.0261, -0.0267, ..., 0.0055, -0.0234, 0.0265], + [-0.0041, -0.0102, -0.0131, ..., 0.0220, 0.0256, -0.0087], + [ 0.0023, 0.0040, -0.0027, ..., -0.0017, 0.0041, 0.0084], + ..., + [ 0.0037, 0.0060, -0.0251, ..., 0.0096, 0.0140, -0.0101], + [-0.0144, 0.0079, 0.0062, ..., 0.0143, 0.0064, -0.0052], + [ 0.0088, -0.0034, -0.0133, ..., 0.0095, 0.0078, 0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_proj.weight': tensor([[-0.0112, 0.0037, -0.0019, ..., -0.0033, -0.0048, -0.0020], + [ 0.0121, 0.0047, 0.0027, ..., -0.0041, -0.0039, 0.0050], + [-0.0042, -0.0073, -0.0038, ..., -0.0074, 0.0123, 0.0027], + ..., + [-0.0063, -0.0048, -0.0037, ..., 0.0152, -0.0041, -0.0047], + [-0.0011, -0.0074, 0.0025, ..., 0.0134, -0.0055, -0.0047], + [ 0.0025, 0.0034, 0.0023, ..., -0.0037, 0.0043, 0.0037]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_norm.weight': tensor([-6.5613e-04, 1.3184e-01, 1.4526e-02, 6.0547e-01, 2.2339e-02, + 1.2634e-02, 8.5449e-03, 3.2806e-03, 7.0190e-03, 1.1279e-01, + 1.8677e-02, 5.0049e-02, 2.6367e-01, 1.3977e-02, 1.7090e-01, + 1.3245e-02, 5.3467e-02, 3.1494e-02, 1.6937e-03, 1.7578e-02, + 8.6060e-03, 1.8262e-01, 2.7344e-01, 1.8555e-01, 2.0752e-02, + 1.6016e-01, 5.5908e-02, 9.7656e-03, 4.0234e-01, 6.2866e-03, + 1.1292e-02, 5.0781e-01, 5.5664e-02, 3.3936e-02, 1.6113e-02, + 1.3123e-02, 6.7871e-02, 4.5508e-01, 5.3906e-01, 1.4954e-02, + 3.8910e-03, 1.0449e-01, 5.2490e-03, 5.3906e-01, 1.8188e-02, + 4.3213e-02, 1.0437e-02, 3.9258e-01, 7.2327e-03, 1.7944e-02, + 6.8359e-02, 1.8799e-02, 1.0437e-02, 5.0391e-01, -8.5831e-05, + 1.9897e-02, 5.5664e-02, 7.7148e-02, 9.0942e-03, 1.0132e-02, + 5.7983e-03, 9.6436e-03, 1.4099e-02, 3.3008e-01, 2.0264e-02, + 2.8534e-03, 2.6001e-02, 5.2979e-02, 6.2988e-02, 1.6235e-02, + 6.1951e-03, 6.8359e-01, -3.4094e-05, 6.1328e-01, 9.0332e-03, + 1.5198e-02, 2.2217e-02, 2.6953e-01, 3.2471e-02, 9.1171e-04, + 1.0645e-01, 9.7656e-03, 4.0771e-02, 1.1169e-02, 2.1362e-02, + 7.7209e-03, 2.9602e-03, 6.9336e-02, 1.9434e-01, 8.9111e-03, + 9.3994e-03, 1.5625e-02, 4.6875e-02, 2.8320e-02, 2.3193e-02, + 2.3560e-02, 1.4709e-02, 1.6846e-02, 6.2012e-02, 1.9287e-02, + 3.1250e-02, 7.2632e-03, 1.0803e-02, 5.4932e-02, 6.5234e-01, + 6.8848e-02, 2.6550e-03, 4.4531e-01, 1.0925e-02, 1.8799e-02, + 1.6479e-02, 1.5015e-02, 8.4043e-06, 8.9844e-02, 4.2969e-01, + 2.4414e-02, 1.7334e-02, 6.2109e-01, 1.1914e-01, 4.0039e-02, + 1.2146e-02, 7.1875e-01, 1.2891e-01, 1.8799e-02, 1.0925e-02, + 4.4250e-03, 2.0752e-02, 1.3046e-03], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_proj.weight': tensor([[-0.0046, -0.0034, 0.0103, ..., 0.0059, -0.0056, 0.0032], + [-0.0042, 0.0003, -0.0034, ..., 0.0005, -0.0081, 0.0029], + [-0.0002, -0.0016, -0.0085, ..., 0.0123, -0.0168, -0.0072], + ..., + [ 0.0040, 0.0046, -0.0043, ..., 0.0052, 0.0020, -0.0061], + [ 0.0093, 0.0029, -0.0031, ..., -0.0035, 0.0093, 0.0023], + [-0.0056, 0.0007, -0.0092, ..., -0.0081, -0.0060, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_norm.weight': tensor([ 6.8665e-04, 1.3184e-01, 1.4038e-02, 6.0156e-01, 2.1851e-02, + 1.2573e-02, 8.7280e-03, 3.2654e-03, 7.4158e-03, 1.1279e-01, + 1.8677e-02, 4.8584e-02, 2.6367e-01, 1.3855e-02, 1.7090e-01, + 1.3916e-02, 5.2734e-02, 3.1738e-02, 1.6327e-03, 1.7456e-02, + 8.9111e-03, 1.8359e-01, 2.7539e-01, 1.8555e-01, 2.0508e-02, + 1.6211e-01, 5.4443e-02, 7.9956e-03, 4.0430e-01, 6.2561e-03, + 1.1536e-02, 5.0781e-01, 5.5664e-02, 3.3936e-02, 1.6479e-02, + 1.2451e-02, 6.8359e-02, 4.5508e-01, 5.3906e-01, 1.4954e-02, + 3.8910e-03, 1.0449e-01, 5.2185e-03, 5.3906e-01, 1.8433e-02, + 4.3213e-02, 1.0803e-02, 3.9258e-01, 7.3853e-03, 1.7822e-02, + 6.8359e-02, 1.9043e-02, 1.0376e-02, 5.0391e-01, 1.2589e-04, + 1.9897e-02, 5.5664e-02, 7.7148e-02, 9.0942e-03, 8.3008e-03, + 5.7373e-03, 9.6436e-03, 1.4404e-02, 3.3008e-01, 1.9775e-02, + 2.9755e-03, 2.5513e-02, 5.2979e-02, 6.2988e-02, 1.6235e-02, + 6.5918e-03, 6.7969e-01, -3.3855e-05, 6.1328e-01, 9.2163e-03, + -1.4771e-02, 2.2217e-02, 2.6953e-01, 3.3203e-02, 9.1934e-04, + 1.0693e-01, 9.6436e-03, 4.0039e-02, 1.0986e-02, 2.0874e-02, + 7.7820e-03, 3.9062e-03, 6.9824e-02, 1.9434e-01, 8.9111e-03, + 9.5825e-03, 1.4709e-02, 4.7119e-02, 2.8442e-02, 2.4048e-02, + 2.3682e-02, 1.4709e-02, 1.6968e-02, 6.2012e-02, 1.9165e-02, + 3.1250e-02, 7.4768e-03, 1.1230e-02, 5.5420e-02, 6.4844e-01, + 6.8848e-02, 2.6855e-03, 4.4727e-01, 1.0620e-02, 1.8677e-02, + 1.6968e-02, 1.5015e-02, 1.1623e-05, 8.9355e-02, 4.2969e-01, + 2.4170e-02, 1.7212e-02, 6.2109e-01, 1.1914e-01, 4.0283e-02, + 1.2085e-02, 7.2266e-01, 1.2793e-01, 1.7822e-02, 1.0681e-02, + 4.3640e-03, 2.0630e-02, 7.3624e-04], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.v_proj.weight': tensor([[ 0.0170, -0.0073, -0.0155, ..., 0.0033, -0.0150, -0.0077], + [-0.0186, -0.0008, -0.0024, ..., -0.0024, -0.0233, 0.0194], + [-0.0188, -0.0067, -0.0031, ..., -0.0038, 0.0283, 0.0312], + ..., + [-0.0130, -0.0038, -0.0086, ..., -0.0001, 0.0154, 0.0029], + [ 0.0229, 0.0045, 0.0222, ..., 0.0052, 0.0056, 0.0077], + [ 0.0139, 0.0012, 0.0033, ..., 0.0089, -0.0164, -0.0210]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0003, 0.0027, ..., -0.0012, 0.0021, 0.0156], + [ 0.0109, -0.0014, 0.0027, ..., 0.0042, 0.0002, -0.0014], + [ 0.0042, 0.0008, -0.0021, ..., -0.0047, -0.0039, -0.0054], + ..., + [-0.0178, -0.0018, 0.0037, ..., 0.0044, -0.0113, 0.0034], + [ 0.0046, -0.0100, 0.0056, ..., -0.0310, 0.0078, -0.0223], + [ 0.0013, -0.0036, 0.0085, ..., 0.0297, -0.0044, 0.0217]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0064, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [-0.0015, 0.0150, -0.0159, ..., -0.0041, -0.0055, 0.0679], + ..., + [-0.0208, 0.0130, -0.0273, ..., 0.0308, -0.0072, -0.0068], + [-0.0337, -0.0025, 0.0054, ..., -0.0040, 0.0172, 0.0231], + [-0.0255, 0.0040, 0.0030, ..., 0.0408, 0.0430, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0031], + [ 0.0198, -0.0208, 0.0130, ..., 0.0089, -0.0337, -0.0055], + [ 0.0302, 0.0101, -0.0023, ..., -0.0304, 0.0199, 0.0024], + ..., + [ 0.0171, -0.0275, -0.0075, ..., 0.0231, 0.0182, -0.0039], + [ 0.0221, 0.0388, 0.0056, ..., 0.0042, -0.0488, -0.0361], + [-0.0005, 0.0120, -0.0051, ..., 0.0037, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img.weight': tensor([[ 0.0273, -0.0041, 0.0118, ..., 0.0012, 0.0347, 0.0615], + [-0.0258, 0.0042, -0.0197, ..., 0.0139, -0.0047, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0097, 0.0132, 0.0184, ..., 0.0110, 0.0087, 0.0515], + [-0.0006, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.2.mlp.layer1.weight': tensor([[ 0.0017, -0.0165, 0.0046, ..., 0.0017, -0.0049, -0.0064], + [-0.0162, 0.0078, -0.0029, ..., 0.0124, 0.0065, -0.0203], + [-0.0122, -0.0107, 0.0084, ..., -0.0070, -0.0010, -0.0610], + ..., + [-0.0042, 0.0038, 0.0038, ..., 0.0156, 0.0085, -0.0038], + [-0.0093, 0.0140, -0.0031, ..., -0.0092, 0.0054, -0.0396], + [-0.0037, 0.0162, -0.0016, ..., 0.0114, 0.0172, -0.0091]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.mlp.layer2.weight': tensor([[ 0.0040, -0.0082, -0.0236, ..., 0.0062, -0.0112, -0.0124], + [-0.0309, -0.0120, -0.0055, ..., -0.0011, -0.0064, -0.0074], + [ 0.0049, -0.0071, -0.0030, ..., 0.0031, -0.0033, 0.0087], + ..., + [-0.0028, -0.0079, -0.0432, ..., -0.0063, -0.0098, -0.0110], + [-0.0104, 0.0050, 0.0120, ..., 0.0024, 0.0128, -0.0052], + [-0.0078, -0.0063, -0.0037, ..., -0.0096, 0.0097, -0.0388]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0396, 0.0398, 0.0381, ..., -0.0042, -0.0042, -0.0044], + [-0.0383, -0.0369, -0.0359, ..., 0.0043, 0.0048, 0.0047], + [-0.0400, -0.0393, -0.0383, ..., 0.0045, 0.0049, 0.0049], + ..., + [-0.0422, -0.0408, -0.0400, ..., 0.0022, 0.0022, 0.0024], + [-0.0415, -0.0408, -0.0398, ..., 0.0021, 0.0022, 0.0022], + [-0.0039, -0.0039, -0.0039, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_self_attn.2.weight': tensor([[ 3.8757e-03, -3.4637e-03, -3.5248e-03, ..., 2.0504e-05, + -5.6744e-05, 5.4016e-03], + [ 3.3722e-03, -4.0283e-03, -3.2654e-03, ..., -4.7302e-03, + -4.6997e-03, -1.7776e-03], + [ 4.8218e-03, -6.3171e-03, -5.7678e-03, ..., -3.7842e-03, + -3.8147e-03, 1.3123e-03], + ..., + [ 3.9062e-02, -4.1992e-02, -4.2480e-02, ..., -2.0630e-02, + -1.7334e-02, 1.6235e-02], + [ 3.9978e-03, -5.7983e-03, -5.8899e-03, ..., 3.8910e-03, + 3.8910e-03, 5.5847e-03], + [-1.4038e-02, 1.5564e-02, 1.4099e-02, ..., 1.7090e-02, + 1.6968e-02, 1.4526e-02]], dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9570e-02, 5.8838e-02, 5.7617e-02, ..., -1.1015e-04, + -3.9339e-05, -1.9169e-04], + [ 6.3965e-02, 6.2012e-02, 5.9326e-02, ..., 3.4904e-04, + 4.6158e-04, 4.9210e-04], + [-4.8340e-02, -4.6387e-02, -4.5654e-02, ..., -2.6464e-05, + 2.8014e-05, -8.0109e-05], + ..., + [ 3.1494e-02, 3.0518e-02, 2.9785e-02, ..., 1.1206e-04, + 4.5061e-05, 7.3433e-05], + [ 6.1035e-02, 5.9326e-02, 5.8594e-02, ..., 5.2643e-04, + 5.6076e-04, 5.5313e-04], + [-4.7852e-02, -4.6143e-02, -4.5166e-02, ..., -3.2234e-04, + -3.6430e-04, -3.2806e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0149, 0.0073, -0.0080, ..., 0.0018, -0.0047, -0.0041], + [-0.0178, 0.0121, 0.0144, ..., -0.0096, -0.0124, 0.0083], + [ 0.0011, 0.0030, -0.0011, ..., 0.0009, 0.0003, -0.0002], + ..., + [ 0.0260, 0.0184, -0.0149, ..., 0.0078, 0.0080, 0.0037], + [ 0.0075, 0.0334, -0.0057, ..., 0.0076, 0.0325, -0.0178], + [-0.0044, 0.0203, 0.0051, ..., -0.0114, -0.0110, 0.0240]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_mlp.1.weight': tensor([[-0.0986, -0.0967, -0.0942, ..., 0.0031, 0.0034, 0.0033], + [-0.0371, -0.0352, -0.0339, ..., 0.0008, 0.0008, 0.0007], + [ 0.0566, 0.0549, 0.0544, ..., -0.0017, -0.0017, -0.0017], + ..., + [ 0.0364, 0.0354, 0.0347, ..., -0.0018, -0.0018, -0.0016], + [ 0.0466, 0.0444, 0.0430, ..., -0.0010, -0.0009, -0.0010], + [-0.0518, -0.0505, -0.0496, ..., 0.0022, 0.0022, 0.0021]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.adaln_modulation_mlp.2.weight': tensor([[-0.0228, -0.0165, 0.0110, ..., -0.0018, 0.0181, -0.0115], + [ 0.0017, -0.0033, -0.0016, ..., -0.0016, 0.0030, 0.0016], + [-0.0005, -0.0007, -0.0008, ..., -0.0010, 0.0007, 0.0009], + ..., + [ 0.0454, -0.0031, -0.0184, ..., -0.0126, 0.0032, 0.0193], + [ 0.0459, -0.0288, -0.0371, ..., -0.0337, 0.0286, 0.0366], + [ 0.0128, -0.0366, -0.0251, ..., -0.0284, 0.0374, 0.0240]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.after_proj.weight': tensor([[-0.0054, -0.0084, -0.0073, ..., 0.0059, 0.0018, -0.0046], + [ 0.0066, 0.0057, 0.0067, ..., 0.0050, 0.0052, 0.0131], + [-0.0096, -0.0041, -0.0051, ..., -0.0023, -0.0037, 0.0110], + ..., + [-0.0015, -0.0067, -0.0140, ..., -0.0055, -0.0112, 0.0187], + [-0.0049, -0.0150, -0.0076, ..., 0.0105, -0.0118, -0.0096], + [-0.0036, 0.0160, 0.0104, ..., -0.0038, 0.0221, -0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_0.2.after_proj.bias': tensor([-3.4790e-03, 2.1820e-03, 1.1139e-03, ..., 4.6387e-03, + 9.3937e-05, -4.8637e-04], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_proj.weight': tensor([[-0.0110, 0.0153, 0.0214, ..., 0.0090, 0.0057, -0.0190], + [ 0.0203, -0.0014, 0.0074, ..., 0.0058, -0.0027, 0.0075], + [-0.0127, -0.0063, 0.0129, ..., -0.0188, -0.0093, -0.0134], + ..., + [ 0.0065, 0.0038, 0.0002, ..., 0.0057, 0.0030, 0.0051], + [ 0.0098, -0.0222, -0.0029, ..., -0.0398, -0.0059, -0.0120], + [-0.0065, -0.0048, -0.0052, ..., 0.0099, 0.0019, 0.0074]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_norm.weight': tensor([ 1.3672e+00, 1.8047e+00, 1.3828e+00, 1.5859e+00, 2.2852e-01, + 5.6396e-02, 6.5308e-03, 4.2915e-06, -1.7694e-33, 3.5477e-04, + 3.9062e-02, 1.2695e-01, 1.0620e-02, 3.2959e-02, -1.1719e-02, + 4.9744e-03, 6.3965e-02, 9.5703e-01, 8.8281e-01, 1.0376e-03, + 8.9169e-05, 1.2734e+00, 2.2188e+00, 1.9609e+00, 1.0469e+00, + 1.1094e+00, 1.5312e+00, 8.9264e-04, -1.6689e-04, -5.9128e-04, + 1.1292e-02, 1.7944e-02, 9.9487e-03, 1.0156e+00, 9.1553e-03, + 1.0156e+00, 7.3242e-04, -6.8665e-05, 7.4768e-03, -1.2024e-02, + 1.9670e-06, -5.9605e-06, 6.8665e-05, 2.3125e+00, 1.5000e+00, + 1.5781e+00, 1.6016e+00, 1.7578e+00, 1.1523e-01, 7.4768e-04, + 2.4048e-02, 2.4707e-01, 1.1172e+00, 1.3281e-01, 1.4038e-03, + 2.1973e-01, 3.0327e-04, 1.2305e-01, 6.1035e-05, -4.4861e-03, + -7.4463e-03, -5.6152e-03, -6.3477e-02, 1.1328e+00, 1.9609e+00, + 8.3984e-01, 1.1094e+00, 1.3281e+00, 7.7637e-02, -2.1484e-02, + 1.2573e-02, 2.5511e-05, -2.9491e-33, -1.1597e-03, 5.2795e-03, + 1.0559e-02, 1.4496e-03, -3.5156e-02, -2.8442e-02, 1.7944e-02, + 8.8867e-02, 3.6133e-02, 1.0254e-02, 6.1951e-03, -3.7994e-03, + 2.8687e-03, 1.0156e+00, 1.1953e+00, 1.6094e+00, 1.5703e+00, + 9.3750e-01, -7.4768e-03, 5.8365e-04, -2.9297e-03, 3.1281e-03, + -6.8665e-03, 2.1729e-02, 5.6152e-02, 1.7090e-02, 4.4678e-02, + -2.8229e-03, -1.3199e-03, 1.5234e-01, 4.2439e-05, 5.4550e-04, + -8.8811e-06, 2.8534e-03, 8.0469e-01, 1.7734e+00, 1.5547e+00, + 1.1016e+00, 1.2031e+00, 6.2988e-02, 6.5994e-04, 4.1992e-02, + 1.2188e+00, 1.1035e-01, 1.1641e+00, -2.0790e-04, 1.3794e-02, + 1.1475e-02, 5.6641e-02, 2.0313e-04, -2.2949e-02, -5.6028e-05, + 1.2695e-02, 1.0234e+00, -9.7656e-03], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.self_attn.k_proj.weight': tensor([[-0.0220, -0.0096, -0.0002, ..., -0.0080, 0.0030, -0.0188], + [ 0.0066, -0.0347, -0.0034, ..., 0.0116, 0.0139, 0.0144], + [-0.0121, 0.0195, -0.0061, ..., -0.0107, -0.0099, 0.0030], + ..., + [ 0.0010, 0.0106, -0.0084, ..., -0.0008, 0.0042, 0.0038], + [ 0.0222, -0.0081, 0.0102, ..., -0.0147, -0.0076, 0.0221], + [-0.0123, -0.0031, -0.0052, ..., 0.0056, 0.0062, 0.0053]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.k_norm.weight': tensor([ 9.6094e-01, 1.7266e+00, 1.3984e+00, 1.4844e+00, 2.3535e-01, + 6.2500e-02, -1.8845e-03, 1.1861e-05, -1.5167e-33, 2.5177e-04, + 3.9551e-02, 1.2793e-01, -1.1780e-02, 5.1025e-02, -1.1902e-02, + -4.7607e-03, 1.0925e-02, 9.5703e-01, 8.8281e-01, -9.4986e-04, + 2.2316e-04, 1.2734e+00, 1.8828e+00, 1.7734e+00, 4.3945e-02, + 5.3906e-01, 1.5469e+00, -6.9580e-03, -3.4714e-04, 4.4823e-04, + -1.1597e-02, 1.7822e-02, -6.1340e-03, 1.0078e+00, -7.6599e-03, + 1.0078e+00, 7.4768e-04, -1.0681e-04, -9.5215e-03, -1.2146e-02, + 2.9206e-06, -7.8082e-06, 6.4373e-05, 1.8672e+00, 1.0234e+00, + 1.5391e+00, 1.6641e+00, 1.7109e+00, 7.5684e-02, 8.0872e-04, + -3.9368e-03, -6.7444e-03, 1.1172e+00, 1.7188e-01, 1.3351e-03, + 2.1973e-01, -8.8501e-04, 1.2305e-01, -1.3173e-05, 7.1289e-02, + -7.4768e-03, 2.4414e-02, -7.1777e-02, 1.1328e+00, 2.0781e+00, + 1.2109e+00, 9.2969e-01, 1.3828e+00, -1.7944e-02, -1.0254e-02, + -1.6968e-02, -2.2888e-05, 4.3815e-33, 1.3275e-03, 1.7456e-02, + 4.6143e-02, 6.2561e-04, 4.9561e-02, -3.9795e-02, -1.7944e-02, + -5.6152e-02, 2.8809e-02, 1.3062e-02, 6.1340e-03, 3.9673e-03, + -1.1215e-03, 1.5234e+00, 1.5625e+00, 1.6953e+00, 1.6172e+00, + 5.4297e-01, 5.5542e-03, -4.3297e-04, -2.9907e-03, 9.4223e-04, + 7.1411e-03, -3.7384e-03, 5.9509e-03, -1.6235e-02, 8.3008e-03, + -2.8076e-03, -1.3275e-03, 1.5234e-01, 3.1233e-05, 5.6839e-04, + -1.0788e-05, 2.8534e-03, 1.6250e+00, 1.9844e+00, 1.5469e+00, + 5.5859e-01, 1.2969e+00, -4.3030e-03, 2.6512e-04, -3.1738e-02, + 1.2188e+00, 1.5442e-02, 1.1641e+00, -2.6512e-04, 1.6724e-02, + 1.1719e-02, 5.5908e-02, 2.0313e-04, 5.7617e-02, -1.0014e-04, + 2.2949e-02, 1.0234e+00, 3.2959e-03], dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.self_attn.v_proj.weight': tensor([[ 0.0020, -0.0152, 0.0064, ..., 0.0461, -0.0211, 0.0038], + [ 0.0127, 0.0267, -0.0182, ..., 0.0046, 0.0138, -0.0134], + [-0.0003, 0.0297, -0.0025, ..., -0.0084, 0.0220, 0.0025], + ..., + [ 0.0086, 0.0102, -0.0201, ..., 0.0080, 0.0078, 0.0047], + [-0.0018, 0.0332, 0.0127, ..., 0.0145, 0.0266, -0.0056], + [-0.0086, -0.0042, 0.0227, ..., 0.0187, 0.0172, -0.0100]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.self_attn.output_proj.weight': tensor([[ 0.0066, 0.0055, 0.0227, ..., 0.0034, 0.0201, -0.0080], + [ 0.0229, -0.0079, 0.0054, ..., -0.0186, 0.0041, -0.0114], + [ 0.0059, 0.0130, -0.0233, ..., -0.0077, -0.0072, 0.0134], + ..., + [-0.0129, -0.0464, -0.0078, ..., -0.0089, -0.0006, -0.0193], + [ 0.0032, 0.0084, -0.0201, ..., 0.0087, 0.0236, 0.0068], + [-0.0566, -0.0126, 0.0322, ..., 0.0216, -0.0221, -0.0166]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_proj.weight': tensor([[ 0.0098, -0.0079, 0.0138, ..., 0.0038, 0.0074, 0.0201], + [ 0.0078, -0.0058, -0.0052, ..., -0.0090, -0.0052, 0.0119], + [-0.0033, 0.0165, -0.0048, ..., 0.0044, -0.0076, -0.0115], + ..., + [ 0.0093, 0.0032, -0.0132, ..., 0.0160, 0.0046, -0.0068], + [-0.0110, -0.0153, -0.0032, ..., -0.0090, 0.0082, 0.0144], + [ 0.0095, -0.0047, -0.0076, ..., 0.0019, 0.0171, -0.0103]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_norm.weight': tensor([0.1064, 0.0159, 0.4277, 0.0698, 0.1680, 0.0361, 0.1484, 0.1357, 0.5938, + 0.0364, 0.0618, 0.0195, 0.0491, 0.0767, 0.0566, 0.1963, 0.4844, 0.0320, + 0.7461, 0.0352, 0.0437, 0.1689, 0.0178, 0.0801, 0.0481, 0.4082, 0.5508, + 0.2559, 0.0771, 0.0535, 0.0403, 0.2158, 0.6680, 0.2441, 0.0518, 0.0181, + 0.0105, 0.1099, 0.0156, 0.0476, 0.0493, 0.1289, 0.0182, 0.0400, 0.2041, + 0.0532, 0.3555, 0.0205, 0.0500, 0.0139, 0.0310, 0.3848, 0.5391, 0.3438, + 0.0903, 0.6484, 0.5977, 0.0400, 0.1235, 0.0322, 0.0854, 0.5586, 0.0102, + 0.5508, 0.1719, 0.3926, 0.0825, 0.0508, 0.0625, 0.4688, 0.0242, 0.5234, + 0.0214, 0.0356, 0.1338, 0.0874, 0.0287, 0.0889, 0.6367, 0.0352, 0.0889, + 0.0825, 0.0435, 0.0391, 0.4219, 0.0640, 0.0625, 0.0986, 0.0352, 0.1426, + 0.0459, 0.1758, 0.0537, 0.0588, 0.0295, 0.1221, 0.0131, 0.3613, 0.5820, + 0.0342, 0.0562, 0.2949, 0.0208, 0.1211, 0.0442, 0.2314, 0.0228, 0.6328, + 0.5859, 0.0327, 0.0120, 0.7695, 0.0192, 0.0137, 0.4941, 0.2070, 0.2754, + 0.7500, 0.0410, 0.0757, 0.0776, 0.5781, 0.6406, 0.0125, 0.0272, 0.2021, + 0.0249, 0.0205], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_proj.weight': tensor([[-8.7280e-03, 1.8158e-03, 7.0496e-03, ..., -5.6458e-04, + 2.1973e-02, -2.3346e-03], + [ 7.2327e-03, -6.7711e-05, -4.7302e-03, ..., -2.6398e-03, + -2.5330e-03, 1.0986e-02], + [-1.3123e-02, 7.6904e-03, -1.5747e-02, ..., 2.2888e-03, + 6.5002e-03, 5.9509e-03], + ..., + [ 2.8839e-03, 3.1128e-03, -2.5330e-03, ..., 2.6855e-03, + -7.0801e-03, 1.2390e-02], + [-1.1780e-02, -1.8539e-03, -1.4465e-02, ..., 5.3406e-03, + -9.0332e-03, 2.7466e-03], + [ 1.0010e-02, -1.4038e-03, 2.1820e-03, ..., -1.3855e-02, + 5.8594e-03, -1.8997e-03]], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_norm.weight': tensor([0.1064, 0.0157, 0.4277, 0.0698, 0.1680, 0.0364, 0.1484, 0.1348, 0.5938, + 0.0369, 0.0623, 0.0195, 0.0491, 0.0767, 0.0569, 0.1982, 0.4844, 0.0320, + 0.7422, 0.0349, 0.0435, 0.1680, 0.0178, 0.0796, 0.0481, 0.4043, 0.5508, + 0.2539, 0.0771, 0.0532, 0.0413, 0.2178, 0.6680, 0.2441, 0.0525, 0.0181, + 0.0102, 0.1099, 0.0159, 0.0466, 0.0488, 0.1299, 0.0183, 0.0403, 0.2041, + 0.0532, 0.3555, 0.0206, 0.0498, 0.0137, 0.0315, 0.3848, 0.5352, 0.3438, + 0.0903, 0.6484, 0.6016, 0.0398, 0.1235, 0.0325, 0.0854, 0.5586, 0.0106, + 0.5508, 0.1709, 0.3926, 0.0825, 0.0505, 0.0625, 0.4707, 0.0240, 0.5234, + 0.0212, 0.0356, 0.1338, 0.0869, 0.0286, 0.0898, 0.6367, 0.0347, 0.0879, + 0.0830, 0.0435, 0.0391, 0.4219, 0.0645, 0.0618, 0.0977, 0.0354, 0.1445, + 0.0469, 0.1748, 0.0537, 0.0596, 0.0298, 0.1216, 0.0131, 0.3613, 0.5820, + 0.0337, 0.0569, 0.2949, 0.0206, 0.1211, 0.0442, 0.2314, 0.0225, 0.6328, + 0.5859, 0.0325, 0.0119, 0.7695, 0.0187, 0.0140, 0.4961, 0.2070, 0.2754, + 0.7461, 0.0403, 0.0732, 0.0771, 0.5781, 0.6406, 0.0123, 0.0272, 0.2021, + 0.0261, 0.0210], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.v_proj.weight': tensor([[ 0.0135, -0.0027, -0.0303, ..., -0.0010, -0.0011, -0.0116], + [-0.0089, 0.0164, -0.0017, ..., 0.0066, -0.0045, -0.0234], + [ 0.0212, 0.0015, -0.0043, ..., 0.0004, -0.0052, -0.0410], + ..., + [-0.0072, 0.0032, 0.0109, ..., -0.0109, -0.0275, 0.0041], + [-0.0396, -0.0017, -0.0281, ..., -0.0083, 0.0090, -0.0190], + [-0.0095, -0.0029, 0.0029, ..., -0.0099, 0.0098, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.output_proj.weight': tensor([[ 0.0021, 0.0048, -0.0103, ..., -0.0038, 0.0046, 0.0111], + [-0.0361, 0.0028, -0.0017, ..., 0.0057, -0.0110, -0.0220], + [ 0.0036, -0.0078, -0.0037, ..., 0.0044, 0.0173, -0.0050], + ..., + [ 0.0079, -0.0135, 0.0210, ..., 0.0182, -0.0076, -0.0128], + [-0.0109, -0.0053, 0.0071, ..., 0.0124, -0.0144, 0.0092], + [-0.0041, -0.0201, 0.0154, ..., -0.0038, 0.0083, 0.0167]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0050, -0.0065, ..., 0.0266, -0.0278, 0.0078], + ..., + [-0.0217, -0.0410, 0.0488, ..., 0.0039, -0.0022, -0.0255], + [-0.0032, 0.0066, 0.0034, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0055, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, -0.0019, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0039, 0.0293, ..., 0.0065, 0.0089, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0140, 0.0126, -0.0093], + ..., + [-0.0044, 0.0320, -0.0233, ..., -0.0067, 0.0038, 0.0187], + [ 0.0136, -0.0352, 0.0267, ..., -0.0305, -0.0039, 0.0005], + [-0.0205, 0.0143, 0.0008, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0014, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0058, -0.0537, -0.0282, ..., -0.0109, -0.0146, -0.0023], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, 0.0029, 0.0015], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0024, ..., -0.0103, 0.0073, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_0.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_0.3.mlp.layer1.weight': tensor([[ 0.0123, 0.0036, 0.0047, ..., 0.0121, -0.0084, -0.0044], + [-0.0125, 0.0001, 0.0205, ..., 0.0079, -0.0267, -0.0065], + [ 0.0048, 0.0181, -0.0098, ..., 0.0228, -0.0131, 0.0217], + ..., + [-0.0101, 0.0042, 0.0571, ..., 0.0067, 0.0117, -0.0193], + [ 0.0087, -0.0047, -0.0078, ..., -0.0101, -0.0037, -0.0166], + [ 0.0046, 0.0154, -0.0159, ..., -0.0304, -0.0219, -0.0162]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.mlp.layer2.weight': tensor([[ 0.0048, 0.0209, -0.0045, ..., -0.0162, -0.0251, 0.0036], + [ 0.0038, 0.0232, 0.0094, ..., -0.0051, 0.0177, 0.0135], + [ 0.0050, -0.0222, 0.0045, ..., -0.0055, -0.0045, 0.0055], + ..., + [-0.0053, 0.0055, 0.0253, ..., -0.0049, -0.0100, 0.0018], + [ 0.0039, 0.0140, 0.0219, ..., -0.0047, 0.0172, 0.0111], + [-0.0066, 0.0059, -0.0039, ..., -0.0128, 0.0103, 0.0061]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0469, -0.0464, -0.0449, ..., 0.0010, 0.0008, 0.0009], + [-0.0386, -0.0381, -0.0371, ..., 0.0027, 0.0027, 0.0027], + [-0.0400, -0.0391, -0.0376, ..., 0.0022, 0.0023, 0.0023], + ..., + [ 0.0381, 0.0371, 0.0359, ..., -0.0036, -0.0037, -0.0036], + [ 0.0327, 0.0317, 0.0306, ..., -0.0031, -0.0030, -0.0030], + [-0.0386, -0.0376, -0.0371, ..., 0.0022, 0.0024, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0152, 0.0145, 0.0137, ..., 0.0220, 0.0248, 0.0148], + [ 0.0016, -0.0010, 0.0036, ..., 0.0025, 0.0036, 0.0003], + [-0.0003, -0.0022, -0.0009, ..., 0.0038, 0.0041, -0.0025], + ..., + [-0.0250, -0.0747, -0.0610, ..., 0.0364, 0.0369, -0.0679], + [ 0.0084, 0.0085, 0.0068, ..., 0.0199, 0.0176, 0.0087], + [ 0.0142, 0.0089, 0.0103, ..., 0.0192, 0.0216, 0.0101]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_cross_attn.1.weight': tensor([[-7.9102e-02, -7.6172e-02, -7.4707e-02, ..., -1.0252e-04, + -1.1730e-04, -1.0729e-04], + [ 3.0029e-02, 2.9297e-02, 2.8809e-02, ..., 9.4995e-07, + 2.6345e-05, 1.4687e-04], + [ 2.8442e-02, 2.7710e-02, 2.6489e-02, ..., 2.7466e-04, + 2.9755e-04, 2.0504e-04], + ..., + [ 6.4453e-02, 6.2988e-02, 6.2012e-02, ..., -1.8978e-04, + -1.8406e-04, -1.1444e-04], + [ 6.6895e-02, 6.4453e-02, 6.2988e-02, ..., -1.7643e-04, + -1.1778e-04, -1.6403e-04], + [-3.6133e-02, -3.5156e-02, -3.4424e-02, ..., -5.2929e-05, + 9.9659e-05, 1.2112e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0038, -0.0112, -0.0028, ..., 0.0201, 0.0288, 0.0041], + [ 0.0071, -0.0090, 0.0116, ..., 0.0005, 0.0060, -0.0019], + [ 0.0011, -0.0026, 0.0036, ..., -0.0082, -0.0055, -0.0093], + ..., + [-0.0141, 0.0030, 0.0330, ..., 0.0052, 0.0139, 0.0052], + [-0.0225, 0.0114, 0.0369, ..., 0.0081, 0.0171, -0.0190], + [ 0.0266, -0.0216, 0.0052, ..., -0.0014, 0.0077, 0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_mlp.1.weight': tensor([[-8.9844e-02, -8.6426e-02, -8.3984e-02, ..., 1.8787e-04, + 3.0899e-04, 2.5368e-04], + [ 2.2461e-02, 2.0264e-02, 1.9409e-02, ..., 1.8787e-04, + 2.6894e-04, 2.6321e-04], + [-9.2285e-02, -9.0332e-02, -8.9844e-02, ..., -4.0245e-04, + -3.2806e-04, -3.6621e-04], + ..., + [-8.2031e-02, -8.0566e-02, -7.9102e-02, ..., 2.1076e-04, + 3.1471e-04, 2.1172e-04], + [-6.1279e-02, -5.9570e-02, -5.8594e-02, ..., 2.2888e-04, + 9.2506e-05, 1.0252e-04], + [ 7.9102e-02, 7.7637e-02, 7.6660e-02, ..., -1.7738e-04, + -2.0599e-04, -1.6880e-04]], dtype=torch.bfloat16), 'net.control_blocks_0.3.adaln_modulation_mlp.2.weight': tensor([[ 0.0023, 0.0038, 0.0038, ..., -0.0055, -0.0086, 0.0069], + [-0.0018, -0.0021, 0.0057, ..., -0.0016, 0.0037, 0.0015], + [-0.0010, 0.0045, -0.0019, ..., -0.0009, -0.0012, 0.0011], + ..., + [-0.0413, 0.0198, -0.0593, ..., -0.0295, -0.0106, 0.0225], + [ 0.0091, -0.0100, 0.0258, ..., 0.0067, 0.0037, -0.0063], + [ 0.0154, 0.0053, 0.0056, ..., 0.0135, 0.0070, -0.0107]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.after_proj.weight': tensor([[-0.0056, 0.0069, 0.0043, ..., 0.0067, 0.0135, -0.0127], + [-0.0050, -0.0069, 0.0068, ..., -0.0057, 0.0049, -0.0186], + [ 0.0104, -0.0056, 0.0124, ..., 0.0151, 0.0187, 0.0063], + ..., + [-0.0077, -0.0034, 0.0146, ..., 0.0031, -0.0150, -0.0096], + [-0.0122, -0.0050, -0.0278, ..., -0.0188, 0.0081, -0.0032], + [-0.0042, 0.0045, 0.0044, ..., -0.0145, 0.0129, 0.0003]], + dtype=torch.bfloat16), 'net.control_blocks_0.3.after_proj.bias': tensor([ 0.0081, -0.0087, -0.0045, ..., -0.0044, -0.0139, -0.0038], + dtype=torch.bfloat16)} +[06-26 07:20:20|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:20|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_0.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:21|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 07:20:21|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 0 checkpoint. +[06-26 07:20:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/blur(ba2f44f2-c726-4fe7-949f-597069d9b91c) +[06-26 07:20:21|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 1 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/eb5325b77d358944da58a690157dd2b8071bbf85/general/blur/ba2f44f2-c726-4fe7-949f-597069d9b91c_ema_bf16.pt +[06-26 07:20:34|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0142, 0.0030, 0.0065], + [-0.0017, 0.0060, -0.0327, ..., -0.0070, 0.0009, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0135, -0.0033, 0.0283], + ..., + [-0.0031, 0.0125, 0.0248, ..., 0.0437, -0.0069, -0.0192], + [ 0.0277, 0.0146, 0.0040, ..., -0.0038, -0.0090, 0.0146], + [ 0.0437, 0.0366, 0.0085, ..., 0.0038, -0.0129, 0.0116]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0045, -0.0310, 0.0300, ..., -0.0068, -0.0048, -0.0339], + [-0.0103, -0.0048, 0.0115, ..., 0.0048, 0.0092, 0.0520], + [-0.0061, 0.0132, 0.0398, ..., -0.0096, -0.0179, -0.0061], + ..., + [-0.0046, 0.0175, -0.0095, ..., 0.0225, 0.0010, 0.0030], + [-0.0209, -0.0081, 0.0087, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0055, 0.0354, -0.0081, ..., 0.0118, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0011, ..., 0.0376, 0.0532, 0.0014], + [ 0.0175, -0.0347, -0.0013, ..., -0.0057, 0.0050, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0154, 0.0262, -0.0112], + [-0.0001, 0.0244, 0.0356, ..., 0.0019, 0.0170, 0.0119], + [ 0.0486, 0.0175, 0.0042, ..., -0.0020, -0.0023, -0.0021]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0141, 0.0569, 0.0115, ..., -0.0063, -0.0337, 0.0076], + [ 0.0491, 0.0413, 0.0251, ..., -0.0053, 0.0393, 0.0047], + [-0.0162, -0.0033, 0.0134, ..., -0.0198, -0.0244, 0.0469], + ..., + [ 0.0014, 0.0261, 0.0022, ..., -0.0041, 0.0347, -0.0033], + [-0.0254, 0.0068, -0.0016, ..., 0.0096, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0122, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0033, -0.0270, -0.0166, ..., 0.0118, -0.0298, 0.0532], + [-0.0031, -0.0035, -0.0128, ..., -0.0161, 0.0149, 0.0308], + [ 0.0144, 0.0086, 0.0146, ..., -0.0118, -0.0410, 0.0354], + ..., + [ 0.0062, -0.0630, -0.0034, ..., 0.0072, -0.0183, 0.0217], + [ 0.0038, 0.0036, 0.0258, ..., -0.0486, -0.0349, -0.0036], + [ 0.0182, -0.0081, 0.0197, ..., 0.0150, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0044, -0.0028, -0.0005, ..., 0.0003, 0.0015, -0.0123], + [-0.0150, 0.0251, 0.0432, ..., -0.0049, 0.0038, -0.0052], + [-0.0153, -0.0203, -0.0245, ..., -0.0120, -0.0325, 0.0040], + ..., + [ 0.0026, 0.0023, 0.0417, ..., 0.0280, 0.0030, -0.0042], + [-0.0024, 0.0001, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0065, -0.0019, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0155, -0.0062, -0.0481, ..., 0.0449, 0.0212, 0.0129], + [ 0.0154, -0.0227, 0.0220, ..., -0.0221, 0.0036, 0.0080], + [-0.0151, -0.0099, 0.0430, ..., -0.0095, 0.0327, -0.0111], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0139], + [-0.0017, 0.0166, -0.0059, ..., 0.0027, 0.0080, -0.0071], + [-0.0110, -0.0028, 0.0117, ..., 0.0058, -0.0034, -0.0075]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0040, -0.0386, -0.0118, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0079, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0126, -0.0044, -0.0126, ..., 0.0216, 0.0103, -0.0055], + ..., + [-0.0152, -0.0172, -0.0137, ..., 0.0098, 0.0098, 0.0184], + [-0.0152, -0.0356, 0.0201, ..., -0.0055, -0.0115, 0.0332], + [-0.0396, -0.0123, -0.0090, ..., -0.0284, -0.0142, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0277, 0.0312, 0.0133, ..., -0.0195, 0.0116, -0.0146], + [ 0.0192, 0.0095, -0.0075, ..., 0.0027, -0.0129, -0.0198], + [-0.0251, 0.0114, -0.0137, ..., 0.0093, 0.0172, 0.0075], + ..., + [ 0.0466, 0.0037, -0.0007, ..., -0.0044, 0.0239, 0.0031], + [ 0.0547, -0.0194, -0.0108, ..., -0.0056, 0.0311, -0.0023], + [ 0.0048, 0.0099, -0.0096, ..., 0.0203, 0.0170, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0045, -0.0032, 0.0283, ..., -0.0110, 0.0031, -0.0182], + [ 0.0049, 0.0129, -0.0315, ..., -0.0457, -0.0063, 0.0106], + [ 0.0027, 0.0261, 0.0245, ..., 0.0046, 0.0080, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0154, 0.0045, -0.0075], + [ 0.0055, -0.0371, 0.0356, ..., -0.0081, 0.0410, -0.0178], + [ 0.0178, -0.0151, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0079, 0.0208, 0.0273, ..., 0.0136, 0.0156, 0.0073], + [-0.0074, 0.0023, 0.0038, ..., 0.0146, 0.0015, 0.0050], + [ 0.0045, -0.0005, 0.0244, ..., -0.0120, -0.0055, -0.0247], + ..., + [ 0.0245, 0.0211, -0.0069, ..., 0.0247, -0.0092, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., -0.0052, -0.0295, 0.0400], + [-0.0116, 0.0132, 0.0305, ..., -0.0108, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0072, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0150, -0.0043, 0.0288, ..., 0.0141, 0.0583, -0.0025], + [-0.0281, 0.0156, -0.0386, ..., -0.0444, -0.0093, -0.0038], + ..., + [ 0.0041, 0.0157, 0.0217, ..., 0.0019, 0.0114, 0.0153], + [ 0.0059, 0.0195, -0.0283, ..., 0.0031, -0.0010, -0.0342], + [-0.0170, -0.0195, -0.0026, ..., 0.0049, 0.0165, 0.0133]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0139, -0.0146, 0.0459, ..., 0.0044, -0.0320, 0.0277], + [ 0.0033, 0.0064, 0.0270, ..., -0.0221, -0.0183, -0.0271], + [ 0.0221, 0.0219, -0.0201, ..., 0.0043, 0.0073, -0.0046], + ..., + [-0.0258, -0.0023, 0.0215, ..., 0.0041, 0.0237, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, 0.0014], + [ 0.0042, -0.0081, -0.0129, ..., -0.0334, 0.0060, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0145, -0.0454, ..., -0.0129, -0.0128, 0.0400], + [ 0.0295, -0.0187, -0.0109, ..., -0.0283, 0.0041, -0.0186], + [ 0.0052, 0.0547, 0.0033, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0280, -0.0058, -0.0269, ..., -0.0087, -0.0287, 0.0148], + [ 0.0091, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0098, 0.0104, ..., -0.0072, 0.0703, -0.0043]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0090, -0.0315, ..., 0.0078, 0.0254, 0.0359], + [ 0.0092, -0.0015, 0.0004, ..., -0.0123, -0.0233, -0.0046], + [-0.0025, 0.0046, -0.0126, ..., -0.0124, -0.0062, 0.0083], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0023], + [ 0.0339, -0.0072, 0.0223, ..., 0.0410, 0.0004, 0.0136], + [-0.0034, -0.0439, -0.0486, ..., 0.0139, -0.0129, 0.0106]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[-0.0035, 0.0141, -0.0068, ..., -0.0052, -0.0043, -0.0229], + [-0.0287, 0.0231, 0.0079, ..., -0.0142, -0.0036, 0.0222], + [ 0.0071, 0.0297, -0.0189, ..., -0.0168, 0.0311, -0.0262], + ..., + [ 0.0233, -0.0049, -0.0050, ..., 0.0045, 0.0278, -0.0034], + [-0.0103, 0.0035, 0.0142, ..., -0.0175, 0.0444, 0.0425], + [ 0.0078, -0.0079, 0.0099, ..., 0.0061, -0.0344, 0.0229]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[-0.0027, 0.0123, 0.0190, ..., -0.0096, -0.0198, 0.0148], + [-0.0581, 0.0294, 0.0032, ..., -0.0240, 0.0071, 0.0302], + [ 0.0087, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0040, -0.0289, 0.0537, ..., 0.0051, -0.0061, -0.0098], + [-0.0041, -0.0081, 0.0209, ..., 0.0049, 0.0101, 0.0186], + [-0.0087, -0.0349, -0.0173, ..., -0.0400, 0.0037, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0181, -0.0265, -0.0108, ..., 0.0096, -0.0133, -0.0405], + [ 0.0078, -0.0090, 0.0266, ..., -0.0400, 0.0073, 0.0391], + [-0.0045, 0.0359, 0.0284, ..., 0.0297, -0.0232, 0.0212], + ..., + [-0.0045, 0.0124, 0.0132, ..., -0.0134, 0.0016, 0.0095], + [ 0.0298, -0.0167, -0.0038, ..., -0.0070, 0.0048, -0.0206], + [ 0.0050, -0.0056, -0.0040, ..., 0.0009, -0.0044, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, 0.0002, 0.0123, ..., 0.0014, -0.0208, -0.0079], + [-0.0096, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0229], + [ 0.0137, -0.0278, 0.0117, ..., 0.0396, -0.0085, -0.0078], + ..., + [-0.0033, 0.0037, -0.0311, ..., 0.0344, 0.0062, -0.0124], + [-0.0236, -0.0100, -0.0067, ..., 0.0332, -0.0117, -0.0173], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0138, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[ 0.0038, 0.0166, 0.0041, ..., 0.0273, -0.0352, 0.0149], + [-0.0510, -0.0076, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0076, ..., -0.0388, 0.0004, 0.0398], + ..., + [-0.0160, -0.0195, -0.0052, ..., -0.0007, -0.0054, -0.0038], + [-0.0187, 0.0040, 0.0094, ..., 0.0287, -0.0122, -0.0098], + [-0.0181, -0.0038, 0.0247, ..., -0.0146, 0.0068, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0059, 0.0097, 0.0140, ..., -0.0112, -0.0019, -0.0253], + [-0.0166, 0.0330, -0.0171, ..., -0.0034, -0.0403, -0.0110], + [ 0.0084, 0.0347, 0.0275, ..., 0.0150, 0.0067, 0.0080], + ..., + [ 0.0271, -0.0167, -0.0106, ..., -0.0098, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0052, ..., 0.0033, 0.0017, -0.0243], + [-0.0079, 0.0581, 0.0067, ..., 0.0147, 0.0239, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0123, -0.0410, 0.0126, ..., -0.0254, -0.0027, -0.0173], + [ 0.0476, 0.0315, 0.0150, ..., -0.0146, -0.0070, 0.0038], + [-0.0092, -0.0270, 0.0086, ..., 0.0332, 0.0148, -0.0075], + ..., + [-0.0067, 0.0028, 0.0162, ..., -0.0177, 0.0171, -0.0134], + [ 0.0048, -0.0187, -0.0052, ..., 0.0248, -0.0140, -0.0058], + [-0.0142, 0.0022, 0.0150, ..., 0.0029, -0.0024, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0063, 0.0093, ..., 0.0255, -0.0253, -0.0132], + [-0.0048, -0.0135, -0.0160, ..., -0.0107, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0172, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0119, 0.0037, 0.0322, ..., -0.0183, -0.0033, 0.0031], + [ 0.0043, 0.0008, 0.0162, ..., 0.0028, 0.0255, 0.0071], + [ 0.0270, 0.0085, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0066, ..., -0.0532, 0.0104, -0.0042], + [-0.0240, -0.0215, -0.0103, ..., -0.0060, 0.0098, -0.0168], + [ 0.0062, -0.0046, -0.0064, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0059, 0.0172, -0.0106, ..., -0.0300, -0.0027, 0.0435], + [ 0.0049, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0066], + [-0.0167, 0.0123, 0.0101, ..., -0.0247, 0.0119, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0072, -0.0079, 0.0317, ..., 0.0199, -0.0105, -0.0376], + [ 0.0356, 0.0378, 0.0134, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., -0.0002, 0.0059, -0.0322], + ..., + [ 0.0059, 0.0107, -0.0040, ..., 0.0366, -0.0206, 0.0374], + [ 0.0024, -0.0079, 0.0040, ..., -0.0099, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0039, ..., -0.0166, -0.0059, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0048, -0.0087, -0.0344, ..., 0.0047, 0.0347, -0.0178], + [ 0.0267, -0.0366, -0.0089, ..., 0.0217, 0.0349, -0.0028], + [-0.0094, -0.0286, 0.0042, ..., -0.0026, -0.0282, 0.0029], + ..., + [-0.0128, 0.0037, 0.0053, ..., -0.0084, 0.0165, 0.0075], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0143], + [-0.0139, 0.0034, -0.0170, ..., -0.0128, 0.0061, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0028, 0.0034, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0118, 0.0374, ..., -0.0153, -0.0188, -0.0136], + ..., + [ 0.0262, -0.0019, -0.0075, ..., -0.0011, 0.0038, -0.0165], + [-0.0222, -0.0154, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0132, -0.0148, 0.0029, ..., 0.0020, -0.0234, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0102], + [ 0.0249, 0.0205, 0.0072, ..., -0.0032, -0.0092, -0.0103], + [-0.0220, -0.0022, -0.0023, ..., 0.0306, 0.0381, -0.0015], + ..., + [ 0.0200, -0.0011, 0.0170, ..., 0.0115, 0.0320, -0.0157], + [ 0.0349, 0.0117, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0067, 0.0256, -0.0023, ..., 0.0205, -0.0045, -0.0106]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0039, 0.0199, ..., 0.0013, 0.0276, -0.0312], + [ 0.0178, 0.0033, 0.0099, ..., 0.0110, -0.0398, 0.0293], + [ 0.0430, -0.0128, 0.0062, ..., 0.0040, -0.0084, 0.0067], + ..., + [ 0.0192, -0.0111, -0.0108, ..., -0.0045, 0.0330, 0.0140], + [-0.0126, -0.0542, -0.0205, ..., 0.0017, -0.0159, 0.0320], + [ 0.0303, 0.0066, -0.0425, ..., -0.0037, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0065, -0.0315, ..., 0.0525, 0.0039, -0.0349], + [-0.0019, -0.0039, -0.0154, ..., 0.0053, -0.0240, 0.0305], + [-0.0256, -0.0028, 0.0072, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0069, 0.0231, 0.0209, ..., 0.0075, 0.0045, 0.0461], + [-0.0072, -0.0079, -0.0001, ..., -0.0217, 0.0347, 0.0039], + [ 0.0032, 0.0234, 0.0038, ..., 0.0261, -0.0033, 0.0100]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0045, 0.0035], + [-0.0183, -0.0211, -0.0037, ..., -0.0145, 0.0138, -0.0162], + [-0.0066, -0.0160, 0.0070, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0096, 0.0078, ..., 0.0171, 0.0040, -0.0053], + [-0.0141, 0.0260, -0.0084, ..., 0.0078, 0.0103, -0.0166], + [ 0.0093, -0.0142, -0.0018, ..., 0.0053, 0.0056, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0070, 0.0141, -0.0053, ..., -0.0278, 0.0240, -0.0115], + [-0.0036, 0.0188, -0.0160, ..., -0.0317, -0.0043, 0.0181], + [-0.0006, 0.0310, -0.0085, ..., -0.0226, 0.0038, 0.0334], + ..., + [-0.0079, 0.0186, -0.0056, ..., 0.0228, 0.0052, 0.0046], + [ 0.0271, 0.0119, 0.0354, ..., 0.0019, 0.0198, 0.0143], + [ 0.0282, 0.0171, 0.0317, ..., -0.0110, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0114, -0.0265, -0.0110, ..., 0.0049, 0.0037, 0.0134], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0081], + [-0.0076, -0.0190, -0.0034, ..., 0.0166, -0.0135, 0.0137], + ..., + [-0.0200, 0.0018, 0.0060, ..., -0.0339, 0.0361, -0.0032], + [-0.0046, -0.0250, 0.0154, ..., -0.0022, -0.0162, 0.0220], + [-0.0044, 0.0038, 0.0205, ..., -0.0064, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0020, 0.0240, -0.0217, ..., 0.0172, -0.0028, 0.0115], + [-0.0386, 0.0103, 0.0074, ..., -0.0041, 0.0322, 0.0087], + [-0.0427, 0.0025, -0.0056, ..., 0.0068, 0.0110, -0.0139], + ..., + [-0.0084, -0.0043, -0.0264, ..., -0.0586, -0.0630, -0.0131], + [ 0.0064, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0106, -0.0610, 0.0162, ..., -0.0291, -0.0166, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0118, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0112], + [ 0.0108, -0.0101, -0.0327, ..., 0.0447, 0.0053, 0.0043], + [-0.0023, -0.0175, 0.0070, ..., -0.0076, -0.0049, 0.0026], + ..., + [ 0.0295, -0.0049, -0.0106, ..., 0.0095, -0.0160, -0.0288], + [ 0.0082, 0.0253, -0.0177, ..., 0.0095, 0.0071, 0.0066], + [-0.0135, -0.0057, 0.0137, ..., -0.0128, 0.0277, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0132, 0.0053, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0016, 0.0234, 0.0062, ..., 0.0464, 0.0069, 0.0309], + ..., + [-0.0118, -0.0175, 0.0327, ..., 0.0132, 0.0014, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, -0.0040, -0.0444, ..., 0.0060, -0.0283, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0081, -0.0101, 0.0146, ..., -0.0742, -0.0077, 0.0161], + [-0.0066, -0.0132, -0.0062, ..., 0.0039, -0.0023, 0.0238], + [ 0.0189, 0.0221, -0.0088, ..., 0.0018, 0.0408, 0.0045], + ..., + [ 0.0046, 0.0247, 0.0396, ..., 0.0043, -0.0304, -0.0183], + [ 0.0299, -0.0112, -0.0417, ..., -0.0039, -0.0376, -0.0045], + [-0.0210, 0.0168, -0.0050, ..., 0.0045, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0044, 0.0242, -0.0042, ..., -0.0245, 0.0033, -0.0070], + [ 0.0190, -0.0121, 0.0332, ..., -0.0144, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0037, ..., -0.0030, -0.0091, 0.0182], + ..., + [-0.0430, -0.0403, -0.0083, ..., 0.0132, -0.0141, 0.0055], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0090, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0050, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0247, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0093, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0061, ..., 0.0101, -0.0223, -0.0012], + ..., + [ 0.0192, 0.0292, -0.0018, ..., 0.0176, -0.0197, 0.0019], + [ 0.0265, 0.0063, 0.0056, ..., 0.0286, 0.0021, -0.0267], + [-0.0184, -0.0088, -0.0045, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0084, -0.0159, 0.0082, ..., 0.0198, 0.0051, 0.0277], + [ 0.0172, 0.0041, -0.0041, ..., -0.0003, 0.0040, 0.0067], + [ 0.0040, 0.0184, 0.0114, ..., 0.0099, -0.0053, -0.0059], + ..., + [ 0.0042, -0.0111, 0.0045, ..., -0.0096, 0.0036, 0.0121], + [-0.0427, -0.0391, -0.0192, ..., 0.0092, -0.0238, -0.0042], + [ 0.0038, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0083, -0.0161, 0.0123], + [ 0.0021, -0.0044, 0.0312, ..., 0.0071, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0071, 0.0046, 0.0278], + ..., + [ 0.0170, -0.0047, 0.0471, ..., -0.0057, 0.0038, -0.0079], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0039, -0.0203], + [ 0.0057, 0.0064, 0.0054, ..., -0.0095, 0.0013, 0.0021]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[ 0.0057, -0.0155, 0.0284, ..., -0.0104, 0.0135, -0.0052], + [ 0.0153, 0.0332, 0.0337, ..., 0.0223, -0.0165, 0.0038], + [-0.0020, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0069, 0.0062, 0.0280, ..., -0.0063, -0.0039, 0.0140], + [-0.0048, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0091, ..., -0.0035, 0.0167, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0045, ..., -0.0238, -0.0102, 0.0119], + [-0.0250, -0.0162, -0.0023, ..., -0.0084, 0.0002, -0.0369], + [ 0.0010, 0.0141, 0.0075, ..., -0.0452, -0.0045, 0.0120], + ..., + [ 0.0147, -0.0141, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0099, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0142, 0.0034, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[-0.0069, -0.0327, -0.0190, ..., 0.0042, 0.0036, -0.0209], + [ 0.0713, 0.0151, -0.0033, ..., 0.0051, 0.0186, 0.0020], + [-0.0227, -0.0092, 0.0188, ..., -0.0122, 0.0193, 0.0075], + ..., + [ 0.0116, 0.0464, 0.0083, ..., 0.0248, 0.0168, -0.0039], + [ 0.0381, -0.0036, 0.0060, ..., 0.0046, -0.0109, 0.0061], + [ 0.0292, 0.0050, 0.0209, ..., -0.0036, 0.0234, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0042, 0.0212, ..., 0.0305, 0.0137, 0.0104], + [-0.0060, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0065], + ..., + [ 0.0084, -0.0275, 0.0225, ..., -0.0039, -0.0189, -0.0107], + [-0.0125, 0.0041, -0.0045, ..., 0.0045, 0.0040, 0.0211], + [ 0.0020, -0.0060, -0.0059, ..., -0.0109, 0.0264, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[ 0.0043, -0.0076, -0.0108, ..., 0.0186, -0.0216, 0.0049], + [ 0.0039, 0.0123, -0.0231, ..., 0.0123, -0.0112, -0.0183], + [ 0.0107, 0.0103, -0.0011, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0123, 0.0146, ..., 0.0540, 0.0415, 0.0097], + [ 0.0038, 0.0087, -0.0138, ..., 0.0055, -0.0189, -0.0237], + [-0.0019, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, 0.0066, 0.0359, ..., -0.0084, -0.0121, -0.0041], + [ 0.0190, -0.0044, 0.0298, ..., 0.0126, 0.0048, -0.0200], + [ 0.0118, -0.0110, 0.0177, ..., -0.0248, 0.0229, -0.0038], + ..., + [-0.0077, -0.0048, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0025, -0.0291, ..., 0.0206, 0.0304, -0.0085], + [-0.0077, 0.0245, -0.0008, ..., -0.0413, 0.0216, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0060, -0.0157, -0.0032, ..., 0.0047, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0072], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, 0.0049], + ..., + [ 0.0033, -0.0058, 0.0258, ..., -0.0310, 0.0092, -0.0361], + [ 0.0040, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [-0.0036, 0.0157, -0.0046, ..., 0.0052, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0070, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0105, ..., 0.0104, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0037, ..., 0.0129, -0.0047, 0.0087], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [ 0.0038, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0128, 0.0206, ..., 0.0058, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0094, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0060, 0.0066, ..., 0.0015, 0.0112, 0.0598], + [-0.0060, 0.0267, -0.0065, ..., 0.0284, -0.0055, -0.0364], + ..., + [-0.0052, 0.0002, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, -0.0036, -0.0089, ..., 0.0211, 0.0134, -0.0078], + [ 0.0102, 0.0236, -0.0206, ..., -0.0134, 0.0048, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0077, 0.0294, ..., 0.0031, 0.0270, 0.0063], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0066, -0.0168], + [ 0.0201, -0.0121, 0.0126, ..., -0.0381, -0.0247, -0.0070], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0078], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0068, -0.0466], + [ 0.0327, 0.0284, -0.0039, ..., 0.0145, 0.0139, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, 0.0039, -0.0167, ..., 0.0005, 0.0170, 0.0107], + [ 0.0038, 0.0009, 0.0295, ..., -0.0172, 0.0038, 0.0135], + [ 0.0193, 0.0236, -0.0125, ..., -0.0045, -0.0369, 0.0239], + ..., + [ 0.0154, -0.0413, -0.0123, ..., -0.0354, 0.0062, -0.0320], + [-0.0266, -0.0071, -0.0327, ..., 0.0040, 0.0049, 0.0043], + [ 0.0298, 0.0356, -0.0192, ..., 0.0055, 0.0039, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0045, 0.0043, 0.0317, ..., 0.0053, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0071, -0.0096], + [-0.0166, 0.0059, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, 0.0049, 0.0260, ..., 0.0159, 0.0054, -0.0474], + [-0.0042, 0.0080, 0.0038, ..., 0.0143, 0.0557, 0.0023], + [-0.0203, 0.0273, 0.0029, ..., -0.0330, 0.0049, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0074, ..., 0.0036, -0.0093, -0.0277], + [ 0.0049, -0.0041, -0.0038, ..., -0.0049, 0.0140, -0.0161], + [-0.0049, 0.0168, -0.0057, ..., -0.0092, 0.0124, 0.0039], + ..., + [ 0.0398, 0.0452, 0.0251, ..., -0.0076, -0.0330, -0.0043], + [-0.0620, 0.0033, 0.0371, ..., 0.0410, 0.0037, -0.0173], + [ 0.0126, -0.0039, -0.0270, ..., -0.0037, 0.0388, -0.0104]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, -0.0039, 0.0245], + [-0.0449, 0.0039, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0061, ..., 0.0029, 0.0288, 0.0124], + [ 0.0330, 0.0040, 0.0059, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0074, -0.0030, ..., -0.0063, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0106, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0045, 0.0029, 0.0527], + [ 0.0041, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0086], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0095, -0.0095, 0.0327], + [ 0.0142, 0.0114, 0.0061, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0126, 0.0281, ..., -0.0242, -0.0126, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0029], + [-0.0053, 0.0039, -0.0286, ..., 0.0054, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0057, ..., -0.0037, 0.0081, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0096, 0.0150], + [-0.0025, 0.0021, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0071, ..., 0.0092, -0.0114, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0070, -0.0017, ..., -0.0220, -0.0145, 0.0040], + [-0.0166, 0.0189, -0.0065, ..., 0.0096, -0.0018, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0006, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0071, ..., -0.0041, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0070, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0122, 0.0289, 0.0125, ..., 0.0134, 0.0078, -0.0103], + [ 0.0188, -0.0197, -0.0039, ..., 0.0284, -0.0342, -0.0072], + [-0.0525, -0.0215, 0.0154, ..., -0.0270, 0.0084, 0.0254], + ..., + [-0.0068, 0.0087, 0.0281, ..., 0.0129, -0.0043, -0.0045], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0038, 0.0026], + [-0.0273, -0.0747, 0.0208, ..., 0.0041, 0.0041, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0047, 0.0150, 0.0334, ..., 0.0205, -0.0273, 0.0040], + [-0.0165, 0.0111, -0.0135, ..., 0.0106, -0.0069, 0.0219], + [ 0.0053, 0.0227, -0.0056, ..., 0.0128, -0.0070, 0.0150], + ..., + [-0.0347, -0.0148, 0.0061, ..., 0.0021, -0.0021, -0.0093], + [-0.0065, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0024, -0.0500, -0.0232, ..., 0.0320, 0.0022, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0059, -0.0364], + [ 0.0220, 0.0036, -0.0114, ..., 0.0047, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, -0.0039, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0074, 0.0192, -0.0393], + [-0.0101, -0.0244, 0.0275, ..., -0.0057, 0.0044, 0.0081], + [-0.0170, -0.0427, 0.0140, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0046, -0.0378, -0.0359, ..., 0.0552, -0.0073, -0.0160], + [ 0.0017, 0.0142, 0.0134, ..., 0.0049, -0.0153, 0.0354], + [ 0.0095, -0.0038, 0.0057, ..., -0.0016, -0.0249, -0.0135], + ..., + [ 0.0078, -0.0126, -0.0059, ..., -0.0095, -0.0084, -0.0028], + [-0.0059, 0.0439, 0.0166, ..., 0.0248, 0.0132, -0.0187], + [ 0.0175, 0.0060, 0.0043, ..., 0.0410, -0.0096, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0107, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0143, -0.0031, ..., -0.0039, -0.0244, 0.0356], + [ 0.0077, 0.0165, -0.0200, ..., -0.0048, 0.0237, -0.0071], + ..., + [-0.0197, -0.0022, 0.0386, ..., 0.0181, 0.0112, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, 0.0035, -0.0374], + [-0.0084, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0047, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0136, 0.0114, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, -0.0032, -0.0040, ..., -0.0048, -0.0030, 0.0337], + ..., + [ 0.0422, 0.0046, 0.0352, ..., -0.0304, -0.0039, -0.0204], + [ 0.0302, 0.0156, 0.0103, ..., -0.0076, -0.0254, -0.0245], + [-0.0041, -0.0069, 0.0042, ..., -0.0045, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, 0.0040, 0.0266, ..., 0.0154, 0.0187, 0.0039], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0041, ..., -0.0172, -0.0091, 0.0231], + ..., + [-0.0295, -0.0137, 0.0001, ..., -0.0041, -0.0052, 0.0099], + [-0.0045, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0075], + [ 0.0040, -0.0048, -0.0259, ..., 0.0045, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[-0.0047, 0.0062, -0.0140, ..., 0.0242, -0.0040, -0.0386], + [-0.0141, 0.0064, -0.0097, ..., -0.0369, -0.0155, 0.0157], + [-0.0267, -0.0020, -0.0068, ..., -0.0127, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0019, 0.0144, ..., 0.0115, 0.0039, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., 0.0041, -0.0039, -0.0137], + [-0.0219, -0.0396, -0.0078, ..., 0.0028, 0.0041, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0036, 0.0168, 0.0194, ..., 0.0349, -0.0420, 0.0013], + [ 0.0043, 0.0273, 0.0146, ..., -0.0184, -0.0035, 0.0154], + [ 0.0145, -0.0039, -0.0078, ..., 0.0245, 0.0359, 0.0059], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0037, 0.0044, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0073, -0.0312, -0.0457], + [ 0.0056, 0.0034, -0.0057, ..., 0.0021, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0037, 0.0121, 0.0039, ..., 0.0126, 0.0018, 0.0074], + [-0.0277, -0.0030, -0.0181, ..., 0.0038, -0.0029, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0043, 0.0513, -0.0113], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0113], + [-0.0282, -0.0042, 0.0121, ..., -0.0045, -0.0017, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0032, -0.0130, ..., 0.0141, 0.0308, 0.0056], + [ 0.0120, -0.0332, -0.0320, ..., -0.0073, -0.0142, -0.0032], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, 0.0045, -0.0311, ..., 0.0120, 0.0058, 0.0082], + [ 0.0043, 0.0198, 0.0141, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0125, -0.0297, -0.0055, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0103, ..., 0.0165, -0.0280, 0.0481], + [-0.0092, 0.0043, 0.0095, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, 0.0019, ..., 0.0302, -0.0025, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0039, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0022, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0048, -0.0222, 0.0262], + ..., + [-0.0147, 0.0049, 0.0260, ..., 0.0044, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0056, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0046, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0147, -0.0249, 0.0114], + [-0.0278, -0.0078, -0.0374, ..., 0.0276, -0.0062, 0.0150], + [ 0.0088, 0.0237, -0.0156, ..., 0.0527, -0.0075, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0072, -0.0095], + [-0.0071, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0034, -0.0129, ..., 0.0039, -0.0371, 0.0043], + [ 0.0059, -0.0127, 0.0043, ..., -0.0204, 0.0039, 0.0065], + [-0.0112, -0.0195, 0.0052, ..., -0.0542, -0.0089, -0.0045], + ..., + [-0.0205, -0.0084, 0.0038, ..., -0.0199, 0.0049, 0.0036], + [-0.0220, -0.0302, -0.0103, ..., 0.0107, 0.0369, -0.0038], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0039, 0.0049, 0.0238, ..., -0.0146, -0.0320, -0.0021], + [-0.0193, 0.0364, -0.0029, ..., -0.0061, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0063, -0.0232, -0.0464], + [-0.0110, -0.0298, -0.0195, ..., 0.0112, 0.0312, 0.0152], + [-0.0070, -0.0119, -0.0195, ..., -0.0042, 0.0109, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0060, -0.0239, -0.0123, ..., -0.0040, 0.0243, -0.0396], + [-0.0100, 0.0153, 0.0061, ..., -0.0277, -0.0045, 0.0016], + ..., + [ 0.0118, 0.0053, 0.0515, ..., 0.0327, -0.0118, -0.0064], + [-0.0040, 0.0170, -0.0006, ..., -0.0066, 0.0280, -0.0210], + [ 0.0012, 0.0075, 0.0041, ..., 0.0154, -0.0061, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0061, 0.0081, -0.0074, ..., -0.0588, 0.0187, -0.0067], + [-0.0103, 0.0199, -0.0078, ..., 0.0080, -0.0332, -0.0043], + ..., + [ 0.0045, -0.0464, 0.0277, ..., -0.0142, 0.0198, -0.0075], + [ 0.0162, 0.0075, 0.0089, ..., -0.0106, 0.0044, 0.0297], + [-0.0270, 0.0132, 0.0166, ..., 0.0081, 0.0175, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0053, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, -0.0007, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, -0.0039, -0.0201, ..., 0.0330, 0.0309, -0.0046], + ..., + [-0.0037, 0.0123, 0.0153, ..., -0.0042, -0.0037, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0039, 0.0175], + [-0.0244, -0.0058, -0.0300, ..., -0.0048, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0017, -0.0058, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, -0.0039, ..., 0.0310, 0.0088, -0.0037], + [ 0.0042, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0109, -0.0237, ..., -0.0104, 0.0157, 0.0064], + [ 0.0079, 0.0205, 0.0146, ..., 0.0262, -0.0195, 0.0082], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0080]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0029, ..., -0.0053, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., 0.0013, 0.0034, -0.0093], + [ 0.0017, 0.0002, 0.0398, ..., -0.0304, -0.0030, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0035, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0040], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0047, 0.0291], + [ 0.0132, -0.0035, -0.0339, ..., 0.0066, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, 0.0036], + [-0.0166, -0.0037, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0062, -0.0039, ..., 0.0240, -0.0037, -0.0031]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0007], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [ 0.0010, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0060, -0.0094], + [ 0.0059, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0078], + [-0.0206, 0.0417, -0.0125, ..., -0.0071, 0.0041, -0.0039], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0041, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, 0.0042, ..., -0.0164, 0.0036, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0028, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0054, -0.0231, -0.0255, ..., -0.0045, 0.0295, -0.0063], + [-0.0255, 0.0513, 0.0039, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0033, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0077], + [-0.0065, -0.0061, -0.0364, ..., -0.0175, 0.0306, -0.0077], + [-0.0051, -0.0195, 0.0019, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0075, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0067, -0.0193, -0.0140], + [-0.0166, -0.0048, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0128, -0.0076, 0.0140, ..., -0.0073, 0.0211, -0.0275], + [-0.0039, -0.0077, -0.0234, ..., 0.0087, -0.0070, -0.0166], + [ 0.0225, -0.0251, 0.0028, ..., 0.0078, 0.0236, 0.0081], + ..., + [ 0.0186, 0.0144, -0.0053, ..., 0.0272, -0.0084, -0.0092], + [-0.0040, -0.0030, 0.0071, ..., -0.0260, 0.0267, 0.0147], + [ 0.0101, 0.0009, 0.0212, ..., 0.0195, -0.0157, 0.0121]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0087, -0.0270, 0.0253, ..., 0.0099, 0.0048, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.1.proj.1.weight': tensor([[-0.0043, 0.0179, 0.0830, ..., -0.0645, 0.0269, -0.0361], + [-0.0288, -0.0048, -0.1138, ..., 0.0220, -0.0299, 0.0554], + [-0.0253, 0.1050, -0.0269, ..., -0.0189, 0.0175, -0.0332], + ..., + [ 0.0144, 0.0107, -0.0055, ..., 0.0226, -0.0234, 0.0046], + [-0.0284, -0.0649, 0.0503, ..., 0.0425, 0.0728, 0.0615], + [ 0.0162, -0.0171, 0.0017, ..., 0.0282, 0.0413, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_proj.weight': tensor([[-0.0054, -0.0018, 0.0101, ..., 0.0036, 0.0021, 0.0046], + [ 0.0051, 0.0034, -0.0007, ..., -0.0087, 0.0004, -0.0032], + [ 0.0102, -0.0041, 0.0022, ..., -0.0049, -0.0089, -0.0043], + ..., + [-0.0117, 0.0017, -0.0068, ..., 0.0032, 0.0063, -0.0047], + [ 0.0031, -0.0067, -0.0085, ..., -0.0039, -0.0055, -0.0053], + [ 0.0129, -0.0121, 0.0019, ..., 0.0051, -0.0067, 0.0019]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_norm.weight': tensor([ 1.5918e-01, -1.1572e-01, 1.1377e-01, 2.5977e-01, 7.2002e-05, + 5.5313e-04, 8.7172e-07, 7.4133e-07, 2.3041e-03, 4.2564e-10, + 1.2878e-09, -4.3165e-13, 1.2719e-12, 3.6948e-13, 4.2810e-13, + 7.7804e-13, 5.3478e-10, -1.0014e-05, -8.0327e-09, 3.6863e-17, + -1.7675e-13, 1.2612e-13, 2.4414e-02, 3.3379e-04, 2.8320e-02, + 6.4453e-01, 1.0400e-01, 4.0283e-03, 5.8838e-02, 9.5215e-02, + -1.3434e-14, -2.3761e-11, 6.0390e-10, -3.6380e-10, 8.9585e-11, + -6.4448e-07, 3.3833e-10, 2.2315e-14, 2.8848e-12, -1.6225e-09, + -5.2296e-12, 7.5033e-11, -5.2736e-15, 9.5367e-05, 2.4536e-02, + 6.1417e-04, -5.3955e-02, 5.8594e-02, 4.5117e-01, -1.0529e-03, + -1.8539e-03, 8.6799e-07, -3.5390e-08, -2.7008e-07, 2.2965e-11, + 9.8225e-10, -2.4556e-10, 1.8813e-07, 2.0227e-09, -2.2055e-11, + -2.7512e-18, -9.1073e-17, -3.8337e-16, -5.5422e-13, 2.7344e-01, + 2.4707e-01, 3.8086e-01, 5.0781e-02, -6.8545e-06, -3.0365e-03, + -6.7521e-09, 3.1710e-05, 2.0294e-03, -2.5437e-12, -1.0959e-10, + 3.0909e-13, 8.7766e-11, 1.1050e-10, 3.5749e-14, 1.7337e-12, + 1.9063e-09, 7.6834e-09, -3.8654e-11, -7.3830e-15, -3.1225e-15, + 1.7167e-11, 3.7537e-03, -1.1673e-03, 6.0156e-01, 7.9102e-02, + 5.3125e-01, 9.1553e-03, 9.8419e-04, 2.2888e-03, 1.9984e-14, + -1.2733e-09, -6.7812e-09, -9.9590e-11, -9.9681e-10, 5.9605e-06, + -3.2152e-13, 4.4409e-16, 2.5757e-14, 2.2847e-09, 2.1982e-14, + 7.9581e-11, -1.2506e-11, 1.7456e-02, -1.8234e-03, -1.6098e-03, + 4.7656e-01, 3.2227e-01, 7.0312e-02, -1.2665e-03, -1.2665e-03, + 1.6928e-05, -3.5288e-10, -7.1526e-07, -1.4097e-10, -6.1846e-10, + -5.4570e-11, 2.3865e-08, 3.2742e-11, -1.5250e-08, 5.2259e-17, + -1.6697e-17, 1.9451e-13, -1.4766e-14], dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.self_attn.k_proj.weight': tensor([[ 0.0060, 0.0034, 0.0004, ..., -0.0048, -0.0052, 0.0039], + [-0.0184, 0.0046, 0.0019, ..., 0.0067, -0.0027, 0.0020], + [ 0.0284, -0.0114, 0.0126, ..., -0.0076, 0.0044, 0.0008], + ..., + [-0.0018, 0.0026, -0.0046, ..., -0.0025, 0.0008, 0.0019], + [-0.0019, 0.0022, -0.0029, ..., -0.0026, 0.0020, 0.0021], + [-0.0021, 0.0010, -0.0025, ..., 0.0011, 0.0010, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.k_norm.weight': tensor([ 2.6758e-01, 1.9043e-01, 1.1426e-01, 2.2461e-01, 4.5538e-05, + -1.8234e-03, -8.0466e-07, -3.5018e-06, 2.4414e-03, 3.0195e-10, + 1.3461e-09, -5.6133e-13, -6.3949e-12, -1.7208e-14, 3.1619e-13, + -7.6739e-13, 6.0754e-10, 1.0133e-05, -6.8976e-09, -8.2399e-18, + 2.0517e-13, 1.1369e-13, 1.9165e-02, -6.6757e-04, 5.0293e-02, + 6.4453e-01, 1.2109e-01, -8.1177e-03, 3.9062e-03, 3.8605e-03, + -7.0083e-16, -2.1737e-10, -2.8667e-09, 3.5652e-10, 5.1841e-11, + 1.2368e-06, 3.3833e-10, -2.3093e-14, -2.9132e-12, 2.0227e-09, + -5.2864e-12, -9.0495e-11, 9.9365e-15, 1.2817e-02, 2.0264e-02, + -7.3433e-05, 1.2451e-01, 1.2061e-01, 4.7656e-01, -1.0376e-03, + 1.9684e-03, -6.7949e-06, 4.1677e-08, -5.0291e-07, 1.8616e-12, + 5.7116e-10, 1.3006e-10, -1.8626e-07, -2.0664e-09, -2.6375e-11, + -3.6863e-18, -2.9317e-16, -3.1398e-16, -5.5778e-13, 8.3496e-02, + 2.5391e-01, 3.8281e-01, -1.7383e-01, 3.1948e-05, -2.0905e-03, + -5.6578e-08, -2.1458e-06, -4.7302e-04, 1.9372e-10, 4.5929e-11, + -2.4514e-13, -7.7307e-11, 1.1914e-10, 6.5725e-14, 9.0949e-13, + -1.7026e-09, 1.0710e-08, -2.9786e-11, 2.7534e-14, 5.1903e-15, + 1.6485e-11, 2.5482e-03, -3.2043e-04, 5.9766e-01, -7.2266e-02, + 5.3125e-01, -4.3640e-03, 5.8105e-02, 9.4238e-02, -2.6201e-14, + 1.2005e-09, 5.5879e-09, 8.9130e-11, 1.0332e-09, -5.3644e-06, + -3.5882e-13, 1.4988e-15, -2.4691e-13, -2.3138e-09, -5.6399e-14, + -7.9581e-11, -1.2449e-11, 1.2390e-02, 1.6479e-02, 1.2360e-03, + 4.7656e-01, 2.9883e-01, 9.6680e-02, 9.9945e-04, -1.9264e-04, + 1.4544e-05, 3.4197e-09, 1.4668e-08, 1.3824e-10, -9.9681e-10, + -1.0823e-10, -2.7358e-08, -5.7128e-12, 1.5483e-08, 5.2692e-17, + -7.6978e-18, 1.9362e-13, -1.5721e-13], dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.self_attn.v_proj.weight': tensor([[ 0.0034, -0.0011, 0.0074, ..., -0.0056, 0.0005, -0.0013], + [ 0.0107, 0.0002, 0.0008, ..., 0.0023, -0.0015, -0.0035], + [ 0.0165, 0.0069, 0.0038, ..., -0.0015, 0.0008, 0.0028], + ..., + [ 0.0009, -0.0049, 0.0047, ..., 0.0081, -0.0025, -0.0019], + [-0.0013, -0.0104, -0.0045, ..., 0.0058, -0.0014, 0.0033], + [-0.0010, 0.0017, -0.0020, ..., -0.0033, 0.0019, -0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.self_attn.output_proj.weight': tensor([[ 6.3171e-03, -2.3071e-02, -2.7222e-02, ..., -3.5858e-04, + -2.0599e-03, 8.5831e-04], + [-3.2043e-03, -8.5449e-03, 1.1353e-02, ..., 2.3499e-03, + 1.0925e-02, -5.2185e-03], + [ 4.8256e-04, 3.1128e-03, -2.5024e-03, ..., -8.8120e-04, + -5.2490e-03, 2.2888e-03], + ..., + [ 3.9673e-03, 5.1270e-03, 7.6294e-03, ..., 1.1444e-03, + -4.1809e-03, -7.2002e-05], + [-1.2085e-02, 7.8125e-03, 8.2397e-03, ..., -2.8687e-03, + 3.0212e-03, 1.2024e-02], + [-6.0425e-03, 4.6997e-03, -9.2773e-03, ..., -1.0071e-02, + 3.3528e-06, 5.1880e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_proj.weight': tensor([[-0.0041, -0.0047, -0.0041, ..., 0.0022, -0.0011, -0.0018], + [-0.0050, -0.0005, 0.0062, ..., 0.0010, -0.0060, -0.0040], + [ 0.0042, 0.0014, 0.0048, ..., -0.0018, 0.0037, 0.0048], + ..., + [-0.0006, 0.0045, -0.0013, ..., 0.0008, -0.0028, 0.0034], + [-0.0029, -0.0054, -0.0033, ..., -0.0016, -0.0026, -0.0128], + [ 0.0023, -0.0041, 0.0075, ..., -0.0037, 0.0009, 0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_norm.weight': tensor([ 1.7700e-03, 1.3447e-04, 1.0885e-08, 1.7319e-03, -2.4319e-05, + 2.3346e-03, -4.5896e-06, 4.3030e-03, 4.0283e-02, 2.7466e-02, + 3.2663e-05, 3.4424e-02, 5.6641e-01, 3.1586e-03, 2.4902e-02, + 1.1921e-04, 4.3750e-01, -1.3947e-05, 2.3041e-03, 5.0659e-03, + 1.1536e-02, 7.9727e-04, 9.2773e-02, 2.3041e-03, 1.0925e-02, + 4.1504e-03, 2.8687e-02, -1.3161e-04, 1.2268e-02, 1.9989e-03, + 3.6133e-01, 8.1787e-03, 2.6226e-05, 4.0820e-01, 5.3516e-01, + 3.3722e-03, 3.8910e-04, 3.8818e-02, 1.6895e-01, 3.1586e-03, + 3.3203e-02, 4.9133e-03, -4.6194e-07, 1.0010e-02, 4.0283e-03, + 3.8574e-02, 1.3794e-02, 1.0620e-02, 3.7689e-03, 9.6798e-05, + 9.5215e-03, 9.5963e-06, 1.9372e-06, 1.7262e-04, 2.2411e-04, + 2.5482e-03, 1.2207e-02, 8.6670e-03, 3.0518e-03, 8.3618e-03, + 5.8365e-04, -1.2779e-04, 1.0925e-02, 9.2163e-03, 2.0905e-03, + 1.2970e-03, 2.8992e-04, 2.6758e-01, 7.8125e-03, 1.6968e-02, + 3.9844e-01, 1.0586e-04, 4.3213e-02, 9.6130e-04, 3.6210e-06, + -1.0757e-03, 5.7068e-03, 8.3542e-04, 5.4932e-04, 1.3000e-02, + 3.7384e-03, 7.1106e-03, 1.8066e-01, 4.0894e-03, 1.0376e-03, + 2.4048e-02, 5.1270e-03, 4.1771e-04, 9.1171e-04, 4.7461e-01, + 4.7302e-03, 2.5330e-03, 1.4587e-02, -7.2956e-05, 2.7275e-04, + 4.2188e-01, 1.1396e-04, 2.2736e-03, -2.8558e-10, 1.1673e-03, + 4.0430e-01, 1.1683e-04, 9.8419e-04, 2.4223e-04, 7.5684e-03, + -1.0252e-04, 1.5106e-03, -2.0409e-04, 3.5248e-03, 5.7373e-03, + 1.1826e-03, 2.3365e-05, 8.4229e-03, 3.7231e-03, 5.0049e-03, + 5.3406e-04, 1.5564e-02, 1.8845e-03, 5.1270e-03, 5.2261e-04, + 7.3433e-05, 7.9956e-03, 5.4932e-03, 2.7847e-04, 1.3123e-03, + 3.4943e-03, 2.0409e-04, 3.8147e-04], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_proj.weight': tensor([[ 1.0452e-03, 7.2861e-04, 2.1057e-03, ..., -8.8120e-04, + -1.5717e-03, -8.1253e-04], + [-3.4790e-03, -2.0599e-03, -4.6997e-03, ..., -2.5787e-03, + -3.6621e-03, -2.4567e-03], + [-1.2512e-03, -1.4877e-03, -2.9907e-03, ..., -2.7466e-03, + -5.3406e-03, -2.0294e-03], + ..., + [ 2.0599e-03, -2.4719e-03, 2.5177e-03, ..., 1.4114e-03, + 1.0633e-04, 2.6703e-04], + [-8.0872e-04, 5.3024e-04, -2.4872e-03, ..., -3.3264e-03, + -5.9605e-05, 1.4877e-03], + [ 4.3106e-04, 1.8997e-03, -2.5024e-03, ..., -4.2114e-03, + 4.4632e-04, 1.0452e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_norm.weight': tensor([ 1.7624e-03, 1.3351e-04, -8.2946e-10, 1.4343e-03, 2.6703e-05, + 2.3499e-03, 5.7817e-06, 4.3030e-03, 3.9307e-02, 2.7466e-02, + 3.1710e-05, 3.4668e-02, 5.7031e-01, 3.1433e-03, 2.4414e-02, + 6.9141e-05, 4.3750e-01, -1.3828e-05, 2.1210e-03, 5.0049e-03, + 1.1536e-02, 8.9264e-04, 9.2773e-02, 2.4414e-03, 1.0864e-02, + 4.0894e-03, 2.8564e-02, -1.3065e-04, 1.1597e-02, 1.9989e-03, + 3.6133e-01, 7.7820e-03, 2.6464e-05, 4.0820e-01, 5.3516e-01, + 3.2654e-03, 5.8365e-04, 3.9062e-02, 1.6992e-01, 3.2959e-03, + 3.3203e-02, 4.8828e-03, -3.3900e-07, 9.8877e-03, 3.4943e-03, + 3.8574e-02, 1.3733e-02, 1.0559e-02, 3.7689e-03, 9.6321e-05, + 9.3994e-03, 9.8348e-06, 3.9339e-06, 1.8501e-04, 1.5869e-03, + 2.5787e-03, 1.2146e-02, 8.6670e-03, 3.5095e-03, 8.3618e-03, + 3.3379e-04, -1.2779e-04, 1.0925e-02, 9.0332e-03, 2.0905e-03, + 1.3351e-03, 2.8992e-04, 2.6758e-01, 7.7820e-03, 1.6968e-02, + 3.9844e-01, 1.0538e-04, 4.1260e-02, 8.8501e-04, 3.7402e-06, + 8.8501e-04, 4.8828e-03, 7.8964e-04, 5.4932e-04, 1.2878e-02, + 3.7689e-03, 7.0801e-03, 1.8066e-01, 3.8910e-03, 9.9945e-04, + 2.3926e-02, 6.3477e-03, 4.1771e-04, 8.8120e-04, 4.7461e-01, + 4.7607e-03, 2.4872e-03, 1.4282e-02, -7.2479e-05, 2.7275e-04, + 4.2188e-01, 6.0797e-05, 2.2583e-03, -3.5288e-10, 1.1597e-03, + 4.0430e-01, 1.2207e-04, 2.1362e-03, 2.3365e-04, 7.5378e-03, + -1.0347e-04, 1.5182e-03, -2.0409e-04, 2.8381e-03, 5.7373e-03, + 1.4877e-03, 3.7670e-05, 8.4229e-03, 3.6926e-03, 5.2185e-03, + -4.1199e-04, 1.5625e-02, 2.0142e-03, 6.0730e-03, 5.5313e-04, + -1.7548e-04, 8.0566e-03, 5.4321e-03, 1.6117e-04, 4.1425e-06, + 3.5095e-03, 2.9945e-04, 2.4319e-04], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.v_proj.weight': tensor([[-1.2390e-02, 3.1586e-03, 1.1902e-02, ..., 3.7842e-03, + 4.3640e-03, -1.0925e-02], + [-2.4658e-02, 2.7120e-06, 8.6060e-03, ..., 1.5030e-03, + -5.3101e-03, 6.1989e-05], + [-1.3000e-02, 3.0670e-03, 5.8594e-03, ..., 2.9602e-03, + 1.4648e-02, -1.2146e-02], + ..., + [ 9.6436e-03, 3.1090e-04, 1.8234e-03, ..., 1.4572e-03, + 5.8899e-03, 1.9043e-02], + [-9.1553e-03, 6.5613e-04, 1.0437e-02, ..., 1.2817e-03, + -1.7090e-02, -3.9368e-03], + [-1.5381e-02, -3.2654e-03, 5.3406e-03, ..., 1.8921e-03, + -8.4229e-03, -2.1362e-02]], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.output_proj.weight': tensor([[ 0.0031, -0.0066, -0.0034, ..., -0.0017, 0.0056, -0.0074], + [ 0.0012, 0.0079, -0.0082, ..., 0.0134, 0.0061, -0.0087], + [-0.0018, -0.0019, -0.0030, ..., 0.0065, -0.0018, -0.0021], + ..., + [ 0.0019, 0.0023, 0.0046, ..., 0.0103, 0.0028, -0.0043], + [ 0.0082, -0.0067, 0.0049, ..., -0.0066, -0.0098, 0.0054], + [ 0.0154, -0.0197, -0.0110, ..., 0.0051, -0.0073, 0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_img.weight': tensor([[ 0.0057, 0.0117, -0.0008, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0077, 0.0249, 0.0291], + [-0.0092, 0.0061, -0.0121, ..., -0.0342, -0.0327, 0.0064], + ..., + [ 0.0070, -0.0040, 0.0016, ..., 0.0085, 0.0286, -0.0165], + [ 0.0071, -0.0060, -0.0100, ..., -0.0098, 0.0143, -0.0309], + [ 0.0242, -0.0104, 0.0255, ..., -0.0228, 0.0039, -0.0156]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0304, ..., -0.0220, 0.0247, 0.0033], + [ 0.0030, 0.0121, 0.0405, ..., 0.0128, -0.0410, 0.0361], + [-0.0089, 0.0113, -0.0087, ..., -0.0057, -0.0038, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0034, 0.0002], + [-0.0160, 0.0047, 0.0013, ..., 0.0134, -0.0175, -0.0044], + [-0.0026, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0102]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0048, -0.0364, ..., -0.0225, -0.0142, -0.0046], + [-0.0157, 0.0058, -0.0364, ..., 0.0099, 0.0119, 0.0079], + [-0.0040, 0.0615, 0.0044, ..., -0.0205, -0.0306, -0.0114], + ..., + [ 0.0038, -0.0040, -0.0103, ..., -0.0211, 0.0014, 0.0288], + [ 0.0266, -0.0117, -0.0052, ..., -0.0420, -0.0282, 0.0048], + [ 0.0286, 0.0325, 0.0097, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.0.mlp.layer1.weight': tensor([[ 5.4932e-03, -5.7373e-03, 4.5776e-03, ..., -4.4250e-03, + 3.2349e-03, -4.3945e-03], + [ 4.7607e-03, -3.5858e-03, -1.3351e-03, ..., 2.7008e-03, + -3.3722e-03, -3.7079e-03], + [ 3.9062e-03, 6.5308e-03, -1.2684e-04, ..., 9.3460e-04, + 2.3041e-03, -3.6469e-03], + ..., + [-5.8289e-03, -6.7444e-03, 3.8719e-04, ..., -9.0332e-03, + 6.6528e-03, -6.7749e-03], + [-2.0294e-03, 8.3618e-03, -5.5237e-03, ..., 6.8054e-03, + 1.2451e-02, -7.4463e-03], + [ 4.2419e-03, -9.6436e-03, -1.7700e-03, ..., 5.3883e-05, + -5.3101e-03, 1.6022e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.mlp.layer2.weight': tensor([[-0.0023, 0.0052, 0.0056, ..., 0.0024, 0.0047, -0.0038], + [-0.0092, 0.0043, 0.0011, ..., -0.0048, 0.0008, 0.0156], + [ 0.0038, 0.0040, -0.0080, ..., 0.0023, -0.0068, 0.0004], + ..., + [ 0.0035, -0.0021, 0.0041, ..., 0.0055, -0.0041, 0.0023], + [-0.0184, 0.0021, 0.0183, ..., -0.0038, 0.0152, -0.0055], + [ 0.0076, -0.0065, -0.0046, ..., 0.0020, 0.0037, -0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0496, -0.0488, -0.0474, ..., 0.0031, 0.0032, 0.0032], + [ 0.0510, 0.0500, 0.0486, ..., -0.0025, -0.0026, -0.0027], + [ 0.0503, 0.0496, 0.0486, ..., -0.0027, -0.0027, -0.0028], + ..., + [ 0.0483, 0.0476, 0.0459, ..., -0.0024, -0.0025, -0.0025], + [ 0.0449, 0.0439, 0.0425, ..., -0.0022, -0.0023, -0.0022], + [ 0.0498, 0.0486, 0.0476, ..., -0.0031, -0.0032, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0046, 0.0041, 0.0043, ..., 0.0056, 0.0057, 0.0046], + [-0.0015, 0.0024, 0.0016, ..., 0.0019, 0.0019, 0.0014], + [ 0.0006, -0.0019, -0.0016, ..., 0.0003, 0.0020, -0.0006], + ..., + [-0.0371, 0.0469, 0.0457, ..., 0.0430, 0.0391, 0.0378], + [-0.0066, 0.0071, 0.0073, ..., 0.0052, 0.0022, 0.0069], + [ 0.0089, -0.0101, -0.0099, ..., -0.0090, -0.0086, -0.0090]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_cross_attn.1.weight': tensor([[-6.6223e-03, -6.2561e-03, -5.4626e-03, ..., -6.3324e-04, + -5.6076e-04, -7.3242e-04], + [-3.2715e-02, -3.1250e-02, -3.1128e-02, ..., 2.0790e-04, + 2.1458e-04, 2.2507e-04], + [ 3.8330e-02, 3.7109e-02, 3.6621e-02, ..., -1.0395e-04, + -8.5831e-05, -1.0967e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1250e-02, ..., -2.8801e-04, + -2.2316e-04, -2.5940e-04], + [ 4.7607e-02, 4.6143e-02, 4.5166e-02, ..., -1.9455e-04, + -1.3447e-04, -1.9360e-04], + [ 3.4424e-02, 3.2959e-02, 3.2715e-02, ..., -1.8358e-05, + 5.8889e-05, 9.4771e-06]], dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_cross_attn.2.weight': tensor([[-1.4526e-02, -4.3335e-03, 3.7079e-03, ..., 4.5776e-03, + 4.6082e-03, 3.4637e-03], + [ 7.0190e-04, -4.8828e-04, 2.1973e-03, ..., 5.8746e-04, + 3.2654e-03, 2.2430e-03], + [ 4.6082e-03, -3.8605e-03, 3.8605e-03, ..., 3.8452e-03, + -7.1049e-05, 3.8605e-03], + ..., + [ 6.1340e-03, -5.5847e-03, 2.4292e-02, ..., 5.5847e-03, + 2.7100e-02, 1.9165e-02], + [ 9.7046e-03, 5.7983e-03, 1.1963e-02, ..., -5.6152e-03, + 1.3046e-03, 1.2085e-02], + [-1.3245e-02, -4.3640e-03, 5.9891e-04, ..., 4.1809e-03, + -1.2817e-03, 1.7262e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_mlp.1.weight': tensor([[-0.0072, -0.0069, -0.0063, ..., 0.0005, 0.0007, 0.0007], + [ 0.0537, 0.0520, 0.0505, ..., -0.0024, -0.0027, -0.0027], + [ 0.0425, 0.0415, 0.0405, ..., -0.0035, -0.0038, -0.0037], + ..., + [ 0.0874, 0.0850, 0.0835, ..., -0.0009, -0.0011, -0.0011], + [ 0.0054, 0.0049, 0.0030, ..., -0.0005, -0.0006, -0.0005], + [ 0.0530, 0.0518, 0.0503, ..., -0.0012, -0.0012, -0.0012]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.adaln_modulation_mlp.2.weight': tensor([[-0.0023, 0.0044, 0.0044, ..., 0.0054, 0.0023, 0.0067], + [-0.0008, 0.0017, 0.0017, ..., 0.0037, 0.0008, -0.0004], + [ 0.0019, 0.0015, 0.0017, ..., 0.0017, -0.0018, -0.0018], + ..., + [-0.0001, 0.0586, 0.0698, ..., 0.0488, 0.0009, 0.0188], + [ 0.0099, 0.0119, 0.0101, ..., 0.0149, -0.0098, 0.0068], + [-0.0082, 0.0038, 0.0038, ..., 0.0038, 0.0081, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.before_proj.weight': tensor([[ 5.5730e-06, -2.1362e-03, 2.8534e-03, ..., -4.4556e-03, + -2.5330e-03, -1.2894e-03], + [ 3.4485e-03, 9.2163e-03, -1.2589e-03, ..., 1.5335e-03, + 3.1586e-03, -3.7994e-03], + [-3.1586e-03, -3.9816e-05, 1.6098e-03, ..., -4.7302e-04, + 4.9210e-04, -1.7624e-03], + ..., + [-5.6839e-04, -1.8387e-03, -3.8757e-03, ..., 9.8419e-04, + 1.6251e-03, -1.9836e-03], + [-8.8215e-05, 6.7749e-03, 1.5411e-03, ..., 2.6131e-04, + 6.3782e-03, 2.7771e-03], + [-6.0120e-03, -5.4932e-03, 1.7700e-03, ..., 6.0654e-04, + 1.2817e-03, -2.1057e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.0.before_proj.bias': tensor([-0.0082, 0.0026, 0.0023, ..., -0.0035, -0.0018, 0.0038], + dtype=torch.bfloat16), 'net.control_blocks_1.0.after_proj.weight': tensor([[-0.0084, 0.0053, 0.0074, ..., -0.0014, 0.0119, 0.0034], + [-0.0170, 0.0038, -0.0052, ..., 0.0042, -0.0137, -0.0132], + [-0.0028, 0.0058, -0.0018, ..., -0.0056, 0.0006, -0.0009], + ..., + [-0.0042, 0.0040, -0.0036, ..., -0.0186, 0.0045, 0.0073], + [ 0.0078, -0.0033, -0.0013, ..., 0.0085, -0.0168, -0.0064], + [ 0.0062, -0.0062, -0.0072, ..., -0.0015, -0.0031, 0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_1.0.after_proj.bias': tensor([ 0.0045, -0.0116, -0.0005, ..., -0.0031, 0.0062, 0.0292], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_proj.weight': tensor([[-3.4424e-02, 9.1553e-03, -9.8267e-03, ..., 1.7334e-02, + 9.5749e-04, -9.5825e-03], + [ 1.3733e-02, -1.9165e-02, 4.2725e-03, ..., -2.0874e-02, + 9.9487e-03, -1.8799e-02], + [-5.3406e-03, -3.6163e-03, -5.1575e-03, ..., -5.1880e-03, + -3.4943e-03, 1.9531e-03], + ..., + [ 2.7924e-03, 3.7384e-03, 2.3193e-03, ..., 4.3030e-03, + 4.5471e-03, 6.9141e-05], + [-3.3264e-03, -3.7231e-03, -1.6403e-03, ..., -3.5706e-03, + 4.1389e-04, -1.7242e-03], + [ 5.6458e-03, -3.7842e-03, 3.2654e-03, ..., -1.7090e-03, + -2.4872e-03, 3.5248e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_norm.weight': tensor([ 2.4512e-01, 8.0859e-01, 1.5430e-01, 1.0469e+00, 8.6719e-01, + 1.5122e-34, 1.9861e-34, -1.9410e-34, 7.5984e-35, 5.8487e-07, + 1.0909e-34, 5.4543e-35, 2.5203e-35, -1.4294e-34, -1.6476e-34, + 9.7049e-35, 1.7676e-01, 2.2888e-04, 2.3079e-04, 9.5215e-03, + 3.4142e-04, 7.1411e-03, 4.1406e-01, 1.0703e+00, 1.1406e+00, + 4.1797e-01, 5.8838e-02, 5.2734e-01, -1.7822e-02, 1.6777e-34, + -4.0324e-34, 1.9034e-34, -3.2806e-03, 9.2697e-04, 4.7684e-04, + 3.9978e-03, 5.1498e-04, -1.5422e-34, -4.9277e-35, 1.3692e-34, + 1.3316e-34, -1.8161e-08, 1.3466e-34, 1.1094e+00, 5.5078e-01, + 4.0039e-01, 3.0664e-01, 1.2422e+00, 1.0547e-01, 1.2360e-03, + 1.1780e-02, 1.9043e-01, 2.6855e-03, -6.9965e-35, -1.6175e-34, + -1.8996e-35, 1.0193e-02, 2.7195e-07, 1.2488e-34, -8.4636e-35, + 3.8556e-35, 3.8744e-35, 1.0608e-34, -2.7836e-35, 5.5078e-01, + 2.7539e-01, 6.7578e-01, 3.6328e-01, 1.3574e-01, -9.5544e-35, + -6.8009e-34, 1.5046e-34, 4.5139e-35, 3.7998e-07, 1.1661e-34, + 1.9259e-34, 2.7836e-35, 7.5608e-35, -1.4219e-34, -1.3090e-34, + -9.3994e-03, 7.6675e-04, 4.4823e-04, 1.6632e-03, 9.3079e-04, + 4.8096e-02, 1.1016e+00, 4.3750e-01, 4.9023e-01, 1.1406e+00, + 2.8516e-01, 2.0117e-01, 6.5308e-03, -5.4769e-34, 2.6933e-34, + 2.0162e-34, 2.4986e-04, -2.4033e-04, 4.7874e-04, 4.5166e-03, + 8.5068e-04, 2.6519e-35, -4.1754e-35, 1.9259e-34, 1.1962e-34, + 1.2293e-07, -1.9259e-34, 3.0859e-01, 1.0625e+00, 1.1250e+00, + 1.2031e+00, 1.7773e-01, 7.6953e-01, 9.1553e-03, -1.9684e-03, + 2.1973e-03, 3.9978e-03, 2.8438e-34, -1.3015e-34, 3.3704e-34, + 9.9945e-04, 1.3828e-05, 6.9589e-35, 3.0394e-34, -1.1285e-35, + -2.1065e-34, 2.1535e-35, 6.3947e-35], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.self_attn.k_proj.weight': tensor([[ 0.0067, -0.0266, 0.0039, ..., -0.0156, -0.0106, -0.0172], + [ 0.0134, -0.0278, 0.0018, ..., -0.0325, 0.0129, -0.0195], + [ 0.0085, 0.0120, -0.0021, ..., -0.0072, -0.0024, -0.0022], + ..., + [ 0.0030, 0.0064, -0.0003, ..., 0.0005, 0.0013, -0.0032], + [-0.0025, -0.0030, -0.0014, ..., -0.0003, 0.0021, -0.0052], + [-0.0003, -0.0024, 0.0031, ..., -0.0048, 0.0014, -0.0027]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.k_norm.weight': tensor([ 4.0234e-01, 8.4766e-01, 2.1777e-01, 1.0625e+00, 8.6719e-01, + 5.9057e-35, 6.3947e-35, 6.7332e-35, -7.4855e-35, -2.3842e-07, + -1.0833e-34, -4.2882e-35, 4.4387e-35, -1.0006e-34, -1.4520e-34, + -1.3843e-34, 1.7676e-01, 3.7766e-04, 6.3896e-05, 4.6387e-02, + 3.3760e-04, -7.2937e-03, 5.1562e-01, 1.0703e+00, 1.1484e+00, + 3.4766e-01, 1.8066e-01, 6.1719e-01, -1.9653e-02, 2.7384e-34, + -4.8524e-35, 1.4896e-34, 3.4027e-03, 9.7656e-04, -1.9169e-04, + 4.6143e-02, 5.2643e-04, -5.0781e-35, -2.4262e-35, 8.5388e-35, + -7.7865e-35, -2.6659e-08, 1.6927e-34, 1.1094e+00, 4.1602e-01, + 1.2012e-01, -2.0605e-01, 1.2266e+00, 1.1377e-01, 1.6479e-03, + 9.2163e-03, 1.2354e-01, 2.1057e-03, 1.0457e-34, 2.0463e-34, + 4.3634e-35, 1.0132e-02, -3.3760e-08, -1.5723e-34, -7.8241e-35, + -1.2413e-35, -6.1314e-35, -1.4294e-34, -2.2569e-35, 4.9023e-01, + 3.3984e-01, 6.7969e-01, 2.5000e-01, 6.2500e-02, 1.5122e-34, + -1.0713e-33, 2.3623e-34, 2.3510e-35, -7.1526e-07, 2.4225e-34, + 2.3623e-34, 1.9184e-35, 8.0874e-35, 2.1535e-35, 1.0382e-34, + 1.7578e-02, 8.2779e-04, 1.6332e-05, 1.4267e-03, 9.3079e-04, + 4.7852e-02, 1.0938e+00, 3.0273e-01, 3.2617e-01, 1.1484e+00, + 2.3926e-01, 1.9434e-01, -2.1362e-04, -1.4520e-34, 2.5127e-34, + -2.9792e-34, -7.9155e-05, 4.1580e-04, 1.8978e-04, 4.5410e-02, + 8.5068e-04, 3.7428e-35, -4.5515e-35, -1.9259e-34, 8.5388e-35, + -1.2480e-07, -2.1516e-34, 3.5156e-01, 1.0781e+00, 1.1328e+00, + 1.2031e+00, -1.1182e-01, 7.8516e-01, 9.2163e-03, 6.6833e-03, + -1.6016e-01, -3.6011e-03, 4.3634e-34, -1.0608e-34, 4.9051e-34, + -1.6403e-03, 1.3828e-05, 2.9716e-35, 1.9184e-34, -3.7616e-35, + -1.0382e-34, -8.4259e-35, -4.7208e-35], dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.self_attn.v_proj.weight': tensor([[ 0.0094, 0.0021, -0.0027, ..., -0.0101, -0.0006, 0.0021], + [ 0.0112, -0.0018, 0.0096, ..., 0.0132, -0.0034, -0.0018], + [ 0.0030, 0.0090, -0.0054, ..., 0.0119, 0.0067, -0.0090], + ..., + [ 0.0101, 0.0050, 0.0096, ..., 0.0025, -0.0022, -0.0150], + [ 0.0035, 0.0030, 0.0087, ..., 0.0302, -0.0042, 0.0026], + [-0.0038, -0.0031, -0.0037, ..., -0.0071, 0.0027, 0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.self_attn.output_proj.weight': tensor([[-0.0118, -0.0044, -0.0154, ..., 0.0201, -0.0093, -0.0011], + [ 0.0154, 0.0047, 0.0190, ..., 0.0036, -0.0079, 0.0216], + [-0.0004, 0.0034, 0.0029, ..., -0.0008, 0.0037, 0.0030], + ..., + [-0.0063, 0.0133, 0.0054, ..., -0.0106, -0.0228, 0.0036], + [ 0.0079, -0.0027, -0.0053, ..., -0.0088, -0.0022, 0.0038], + [ 0.0028, 0.0053, -0.0009, ..., 0.0108, 0.0050, 0.0019]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_proj.weight': tensor([[-0.0126, 0.0129, 0.0119, ..., 0.0013, 0.0060, -0.0038], + [ 0.0101, -0.0017, 0.0041, ..., 0.0045, -0.0068, -0.0172], + [ 0.0040, 0.0110, -0.0029, ..., 0.0031, 0.0066, -0.0067], + ..., + [ 0.0084, -0.0129, 0.0032, ..., -0.0044, -0.0045, 0.0088], + [-0.0099, 0.0049, 0.0044, ..., 0.0073, 0.0020, 0.0093], + [ 0.0398, -0.0062, -0.0023, ..., 0.0098, 0.0096, -0.0177]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_norm.weight': tensor([ 2.2217e-02, 4.4922e-01, 2.2583e-02, 9.9487e-03, 4.8218e-03, + 8.4639e-06, 7.0572e-04, 5.1562e-01, 7.6675e-04, 3.9062e-01, + 2.9785e-02, 1.9653e-02, 8.4229e-03, 1.5015e-02, 1.8921e-02, + 1.4648e-02, 8.5449e-03, -9.6798e-05, 6.1512e-05, 3.1738e-02, + 2.6172e-01, 2.5635e-02, 5.4016e-03, 1.2512e-02, 6.2561e-03, + 1.2302e-04, -2.0266e-05, 4.8584e-02, 7.5102e-06, 3.7109e-01, + 2.0790e-04, 7.8082e-06, 9.6436e-03, -3.2336e-06, 1.0559e-02, + 7.2021e-03, 6.2466e-05, 3.9844e-01, 1.4343e-02, 6.5613e-03, + 7.0801e-02, 6.4392e-03, 1.0986e-02, 4.5471e-03, 1.2817e-02, + 8.6060e-03, 6.3171e-03, 1.7822e-02, 2.4902e-02, 1.3062e-02, + 1.4160e-02, 4.0234e-01, 1.1169e-02, 1.9897e-02, 3.1982e-02, + 1.3351e-03, 9.6680e-02, 1.3611e-02, 2.0264e-02, 1.4954e-02, + -5.2750e-06, 1.9043e-02, 1.5625e-02, 1.3053e-05, 1.2268e-02, + 8.2397e-03, 1.4496e-04, 8.3923e-04, 1.6357e-02, 7.8125e-02, + 1.2131e-03, 1.3123e-02, 4.8828e-02, 3.9795e-02, 1.6327e-03, + 1.8433e-02, -1.6212e-05, 2.5269e-02, 2.5635e-02, 2.5269e-02, + 6.8970e-03, 1.8555e-02, 3.4424e-02, -3.5286e-05, 2.1240e-02, + 3.3569e-03, 1.5030e-03, 4.4189e-02, 6.3782e-03, 1.4267e-03, + 1.6602e-02, 2.2030e-04, 7.4768e-03, 2.0020e-02, 9.0942e-03, + 5.9128e-05, 1.7456e-02, 4.4189e-02, 1.5991e-02, 4.2236e-02, + 3.8330e-02, 2.7588e-02, 1.2878e-02, 9.3460e-05, -3.5286e-05, + 5.5847e-03, 1.1414e-02, 2.1484e-02, 1.7456e-02, 2.5269e-02, + 7.3547e-03, 6.4087e-03, 3.5400e-02, 4.9023e-01, 3.0212e-03, + 1.6724e-02, 6.8054e-03, 1.1292e-02, 7.5989e-03, 1.0547e-01, + 1.4648e-03, 1.3733e-03, -9.6560e-06, 1.2634e-02, 5.9204e-03, + 2.1851e-02, 2.7466e-02, 1.6846e-02], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_proj.weight': tensor([[-0.0032, -0.0033, -0.0042, ..., -0.0037, -0.0011, 0.0017], + [ 0.0116, -0.0035, 0.0184, ..., 0.0208, 0.0198, 0.0116], + [-0.0008, 0.0036, 0.0039, ..., 0.0021, 0.0052, -0.0018], + ..., + [-0.0070, -0.0049, -0.0031, ..., -0.0012, 0.0021, -0.0016], + [-0.0072, -0.0053, -0.0142, ..., 0.0007, 0.0086, -0.0025], + [ 0.0007, -0.0004, -0.0055, ..., -0.0012, 0.0014, -0.0043]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_norm.weight': tensor([ 2.2827e-02, 4.4922e-01, 2.2461e-02, 9.9487e-03, 4.8523e-03, + -6.1095e-06, 7.2098e-04, 5.1562e-01, 8.0109e-04, 3.9062e-01, + 3.0518e-02, 1.9653e-02, 8.4229e-03, 1.5015e-02, 1.8921e-02, + 1.4526e-02, 9.0942e-03, -9.7275e-05, 5.2214e-05, 3.1982e-02, + 2.6172e-01, 2.4902e-02, 5.4321e-03, 1.2756e-02, 6.1951e-03, + 1.2493e-04, -1.8597e-05, 4.8096e-02, 6.1989e-06, 3.7109e-01, + 2.0885e-04, 9.1791e-06, 9.3384e-03, -1.7643e-05, 1.0559e-02, + 7.2632e-03, 6.2466e-05, 3.9844e-01, 1.4404e-02, 6.5002e-03, + 6.9824e-02, 6.4087e-03, 1.0803e-02, 4.5471e-03, 1.2756e-02, + 8.5449e-03, 6.4087e-03, 1.7700e-02, 2.4780e-02, 1.3306e-02, + 1.4160e-02, 4.0234e-01, 1.1169e-02, 1.9531e-02, 3.1982e-02, + 1.3275e-03, 9.6680e-02, 1.3611e-02, 1.9897e-02, 1.5076e-02, + -7.5102e-06, 1.9043e-02, 1.5625e-02, 1.3053e-05, 1.2207e-02, + 8.1177e-03, 1.4400e-04, 8.3542e-04, 1.6357e-02, 7.8125e-02, + 1.1597e-03, 1.3123e-02, 4.9072e-02, 3.9795e-02, 1.6632e-03, + 1.8555e-02, -2.3246e-05, 2.5024e-02, 2.5757e-02, 2.5269e-02, + 6.8970e-03, 1.8433e-02, 3.3936e-02, -3.6001e-05, 2.1118e-02, + 3.2501e-03, 1.4801e-03, 4.3945e-02, 6.6528e-03, 1.7242e-03, + 1.6968e-02, 2.5558e-04, 7.4158e-03, 1.9897e-02, 8.9111e-03, + 5.9128e-05, 1.6968e-02, 4.3701e-02, 1.5991e-02, 4.1992e-02, + 3.9062e-02, 2.7588e-02, 1.3000e-02, 9.0122e-05, -3.5524e-05, + 5.5542e-03, 1.1475e-02, 2.1240e-02, 1.7456e-02, 2.5146e-02, + 7.3547e-03, 6.1646e-03, 3.4912e-02, 4.8828e-01, 3.0670e-03, + 1.6724e-02, 6.8054e-03, 1.1169e-02, 7.7820e-03, 1.0547e-01, + 1.5488e-03, 1.3199e-03, -5.1856e-06, 1.3000e-02, 5.9814e-03, + 2.1973e-02, 2.7466e-02, 1.6724e-02], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.v_proj.weight': tensor([[-0.0178, -0.0001, -0.0044, ..., 0.0012, -0.0193, -0.0247], + [-0.0055, 0.0055, 0.0051, ..., 0.0037, 0.0095, 0.0070], + [-0.0400, -0.0034, -0.0039, ..., -0.0058, -0.0315, -0.0082], + ..., + [-0.0417, 0.0030, -0.0164, ..., 0.0028, 0.0205, 0.0055], + [ 0.0082, 0.0029, -0.0052, ..., 0.0016, 0.0115, 0.0044], + [ 0.0112, 0.0037, -0.0070, ..., 0.0033, 0.0131, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.output_proj.weight': tensor([[ 0.0064, 0.0042, 0.0162, ..., -0.0038, 0.0042, 0.0040], + [-0.0072, 0.0031, -0.0103, ..., -0.0123, 0.0031, 0.0117], + [-0.0011, 0.0020, -0.0045, ..., 0.0019, 0.0011, -0.0018], + ..., + [-0.0021, -0.0033, 0.0043, ..., -0.0018, 0.0036, 0.0046], + [ 0.0087, 0.0232, 0.0022, ..., -0.0349, 0.0075, 0.0289], + [ 0.0039, 0.0022, -0.0156, ..., -0.0105, 0.0062, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_img.weight': tensor([[ 0.0079, -0.0041, 0.0430, ..., -0.0327, 0.0048, 0.0144], + [-0.0278, 0.0280, 0.0249, ..., -0.0077, -0.0378, 0.0271], + [ 0.0139, -0.0119, 0.0312, ..., -0.0072, 0.0253, -0.0111], + ..., + [ 0.0245, -0.0152, 0.0043, ..., -0.0452, 0.0216, -0.0209], + [-0.0089, 0.0076, -0.0471, ..., -0.0107, 0.0157, -0.0083], + [ 0.0074, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0027, -0.0122, -0.0108], + [ 0.0135, 0.0048, -0.0188, ..., 0.0156, 0.0074, -0.0045], + [-0.0306, -0.0021, -0.0272, ..., 0.0103, 0.0337, 0.0017], + ..., + [-0.0034, -0.0042, 0.0049, ..., 0.0208, -0.0107, 0.0065], + [ 0.0193, 0.0210, 0.0122, ..., -0.0145, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0106, -0.0132, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img.weight': tensor([[ 0.0038, -0.0043, 0.0173, ..., 0.0083, 0.0356, 0.0217], + [-0.0498, -0.0033, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0108, 0.0043, -0.0139, ..., 0.0007, 0.0269, 0.0072], + ..., + [-0.0043, 0.0126, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0302, 0.0080, -0.0162, ..., -0.0050, -0.0074, -0.0118], + [ 0.0254, 0.0130, -0.0425, ..., -0.0264, 0.0138, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.1.mlp.layer1.weight': tensor([[ 0.0022, -0.0042, 0.0031, ..., 0.0037, 0.0016, -0.0054], + [-0.0031, 0.0053, 0.0071, ..., -0.0023, -0.0075, 0.0022], + [ 0.0026, 0.0027, 0.0036, ..., -0.0020, -0.0065, 0.0103], + ..., + [ 0.0281, 0.0128, -0.0095, ..., 0.0020, -0.0066, 0.0111], + [-0.0449, 0.0030, -0.0078, ..., 0.0183, 0.0075, -0.0037], + [ 0.0094, -0.0079, 0.0070, ..., -0.0075, 0.0014, -0.0247]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.mlp.layer2.weight': tensor([[-0.0096, -0.0112, 0.0197, ..., 0.0036, -0.0042, -0.0118], + [-0.0065, 0.0038, -0.0004, ..., -0.0036, -0.0001, -0.0032], + [ 0.0049, 0.0027, -0.0094, ..., -0.0126, 0.0036, 0.0025], + ..., + [ 0.0063, -0.0124, 0.0052, ..., -0.0036, -0.0082, 0.0097], + [ 0.0051, 0.0172, 0.0078, ..., 0.0063, 0.0067, 0.0023], + [ 0.0024, -0.0042, 0.0013, ..., -0.0013, -0.0005, 0.0079]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0410, -0.0398, -0.0393, ..., 0.0050, 0.0053, 0.0052], + [ 0.0327, 0.0315, 0.0305, ..., -0.0039, -0.0038, -0.0039], + [-0.0449, -0.0435, -0.0430, ..., 0.0051, 0.0052, 0.0052], + ..., + [-0.0454, -0.0442, -0.0430, ..., 0.0034, 0.0035, 0.0036], + [-0.0403, -0.0396, -0.0388, ..., 0.0037, 0.0039, 0.0039], + [-0.0231, -0.0222, -0.0211, ..., 0.0014, 0.0014, 0.0015]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_self_attn.2.weight': tensor([[ 6.8359e-03, -1.9302e-03, 5.8594e-03, ..., 6.8054e-03, + 6.8970e-03, 2.0142e-03], + [-1.6174e-03, 1.6403e-03, -1.7319e-03, ..., -2.3956e-03, + -1.7853e-03, -7.0572e-04], + [-1.0864e-02, -5.0735e-04, -9.8877e-03, ..., -1.1292e-02, + -1.0803e-02, 2.0862e-05], + ..., + [-1.8433e-02, -4.3678e-04, -2.4414e-02, ..., -2.6611e-02, + -2.5757e-02, 8.3542e-04], + [-1.0986e-02, -6.4697e-03, -1.1780e-02, ..., -1.1780e-02, + -1.0376e-02, 7.3242e-03], + [ 1.7395e-03, -6.0730e-03, 2.8534e-03, ..., 2.6550e-03, + 2.0447e-03, 5.0049e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5176e-02, 5.3711e-02, 5.3223e-02, ..., -1.0872e-04, + -1.5354e-04, -1.0729e-04], + [ 4.2114e-03, 4.1504e-03, 4.1504e-03, ..., -7.5340e-05, + -1.5163e-04, -2.6703e-04], + [-1.7700e-02, -1.6968e-02, -1.6235e-02, ..., -6.6757e-04, + -6.6757e-04, -6.7902e-04], + ..., + [-2.6001e-02, -2.6001e-02, -2.5024e-02, ..., -4.2343e-04, + -3.8147e-04, -3.4332e-04], + [-2.7100e-02, -2.6733e-02, -2.5635e-02, ..., -2.1839e-04, + -1.7262e-04, -2.2602e-04], + [ 5.4199e-02, 5.3223e-02, 5.2246e-02, ..., -1.7262e-04, + -5.4836e-05, -2.0218e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_cross_attn.2.weight': tensor([[-8.9111e-03, -4.9133e-03, -1.1353e-02, ..., -6.1646e-03, + 9.2163e-03, -9.5825e-03], + [ 9.6436e-03, -1.9150e-03, -3.1128e-03, ..., 2.9755e-03, + -8.4229e-03, 1.0071e-02], + [ 5.7983e-03, 1.8215e-04, -2.6941e-05, ..., 3.1281e-03, + -4.3335e-03, 5.3101e-03], + ..., + [ 6.0547e-02, 7.0190e-03, -4.3030e-03, ..., 5.7678e-03, + -1.6113e-02, 6.0059e-02], + [ 1.0620e-02, -4.6082e-03, -8.7891e-03, ..., 4.1504e-03, + -1.5625e-02, 1.1230e-02], + [-3.8605e-03, 9.1553e-03, -1.5503e-02, ..., 1.0376e-02, + 7.3242e-03, -3.8452e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_mlp.1.weight': tensor([[-2.8687e-02, -2.8564e-02, -2.8198e-02, ..., 3.9816e-05, + -4.6194e-06, 1.8477e-05], + [ 3.6621e-02, 3.6133e-02, 3.5889e-02, ..., -3.5248e-03, + -3.5553e-03, -3.4790e-03], + [ 3.9795e-02, 3.9062e-02, 3.8574e-02, ..., -3.6621e-03, + -3.6621e-03, -3.7079e-03], + ..., + [ 4.1016e-02, 3.9795e-02, 3.9551e-02, ..., -1.8616e-03, + -1.8845e-03, -1.8234e-03], + [ 1.9653e-02, 2.0264e-02, 2.0264e-02, ..., 1.6937e-03, + 1.6708e-03, 1.5640e-03], + [ 3.4912e-02, 3.4912e-02, 3.4668e-02, ..., -2.2583e-03, + -2.3193e-03, -2.3346e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.adaln_modulation_mlp.2.weight': tensor([[-0.0033, 0.0037, -0.0026, ..., 0.0036, 0.0035, 0.0038], + [ 0.0029, -0.0019, -0.0007, ..., -0.0013, -0.0033, -0.0018], + [ 0.0055, -0.0088, -0.0107, ..., -0.0078, -0.0045, -0.0078], + ..., + [ 0.0024, -0.0120, -0.0199, ..., -0.0101, 0.0016, -0.0090], + [ 0.0038, -0.0038, -0.0087, ..., -0.0038, -0.0038, -0.0038], + [ 0.0086, -0.0053, -0.0040, ..., -0.0063, -0.0093, -0.0060]], + dtype=torch.bfloat16), 'net.control_blocks_1.1.after_proj.weight': tensor([[-3.2227e-02, 6.6528e-03, 3.7384e-03, ..., -1.4465e-02, + 7.3242e-03, -5.8899e-03], + [ 9.6436e-03, -6.9141e-05, 6.8054e-03, ..., 1.1902e-02, + 6.5918e-03, 1.8539e-03], + [ 1.4771e-02, -7.2021e-03, -1.2085e-02, ..., -4.2419e-03, + 5.5847e-03, 4.0894e-03], + ..., + [-7.9956e-03, 6.7139e-03, 4.1504e-03, ..., 1.5198e-02, + 6.6223e-03, 6.1340e-03], + [-1.2024e-02, -2.3499e-03, 5.7983e-03, ..., 5.2795e-03, + 6.1646e-03, 6.5613e-03], + [ 5.5847e-03, -7.5684e-03, 5.6458e-03, ..., 5.3406e-03, + -6.8359e-03, -1.2894e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.1.after_proj.bias': tensor([-0.0041, -0.0006, 0.0020, ..., -0.0030, -0.0014, -0.0011], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_proj.weight': tensor([[ 0.0452, -0.0189, -0.0077, ..., 0.0088, -0.0152, 0.0352], + [-0.0103, -0.0201, -0.0019, ..., -0.0052, 0.0085, -0.0058], + [ 0.0259, 0.0054, -0.0006, ..., 0.0060, -0.0057, 0.0159], + ..., + [-0.0143, -0.0048, 0.0012, ..., 0.0034, -0.0030, -0.0065], + [ 0.0223, 0.0277, 0.0019, ..., 0.0176, -0.0047, 0.0086], + [-0.0100, 0.0103, 0.0188, ..., -0.0136, -0.0041, -0.0189]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_norm.weight': tensor([ 1.4844e+00, 3.2617e-01, 4.1602e-01, 3.4570e-01, 1.3750e+00, + 2.7466e-04, 1.0437e-02, 5.0962e-06, 1.9775e-02, 5.3024e-04, + -9.7789e-08, -2.5705e-07, 1.1484e+00, 3.9978e-03, -3.8369e-13, + -1.6479e-03, -1.9379e-03, -2.3499e-03, 3.0396e-02, 1.5869e-02, + 1.5820e-01, -8.8692e-05, 4.2578e-01, 1.3594e+00, 5.0000e-01, + 5.5078e-01, 3.8477e-01, 1.6699e-01, 2.5513e-02, -2.5482e-03, + 1.3611e-02, -7.2098e-04, 1.4160e-02, 5.8899e-03, -1.2512e-03, + 3.4180e-02, 1.2988e-01, 5.9891e-04, -2.2430e-03, -4.0829e-06, + -1.2589e-04, 1.5182e-03, 2.2888e-03, 1.1562e+00, 9.0332e-02, + 1.3828e+00, 5.8203e-01, 2.0117e-01, -3.6865e-02, 4.9744e-03, + -2.3651e-03, 3.0273e-02, 1.3086e-01, 5.3048e-06, -7.1526e-05, + -1.1176e-06, -2.0981e-04, -6.3419e-05, -4.4678e-02, 4.8021e-09, + 3.5156e-02, -1.7262e-04, -2.9541e-02, -2.0020e-02, 6.0156e-01, + 1.4297e+00, 1.3281e+00, 1.3984e+00, 2.9688e-01, -1.6846e-02, + 1.3245e-02, 7.3016e-06, 2.1118e-02, -1.3367e-02, 1.4808e-07, + -2.3395e-06, 5.8105e-02, 2.7618e-03, 2.0073e-13, 8.1250e-01, + 1.0625e+00, 1.3000e-02, 1.9550e-04, 3.1982e-02, 1.1230e-01, + 9.8267e-03, 1.3359e+00, 3.1250e-01, 1.3672e+00, 1.4844e+00, + 2.2266e-01, 4.8047e-01, 4.7852e-02, -2.0630e-02, -5.2185e-03, + -6.7234e-05, -9.0332e-03, 1.4832e-02, -2.1515e-03, 3.3447e-02, + 1.2158e-01, 6.1989e-05, 1.1562e+00, 6.1095e-07, -4.6015e-05, + -3.7842e-03, 3.0100e-06, 7.1484e-01, 1.4219e+00, 1.3672e+00, + 1.5781e+00, 5.2734e-01, 2.7275e-04, 5.5542e-03, 6.3171e-03, + 3.1250e-02, 1.3770e-01, 1.4365e-05, -2.9504e-06, -6.2957e-07, + -6.6757e-05, -1.0452e-03, 1.1250e+00, 4.5262e-07, 2.8442e-02, + -1.5378e-05, 7.8125e-01, 1.1016e+00], dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.self_attn.k_proj.weight': tensor([[-0.0225, 0.0005, 0.0068, ..., 0.0205, 0.0119, 0.0415], + [-0.0079, 0.0108, -0.0090, ..., -0.0039, -0.0042, -0.0036], + [-0.0143, 0.0108, -0.0109, ..., 0.0058, 0.0182, 0.0007], + ..., + [-0.0030, -0.0069, 0.0044, ..., 0.0017, 0.0044, 0.0078], + [-0.0130, -0.0059, 0.0067, ..., 0.0051, 0.0056, 0.0075], + [ 0.0010, -0.0042, 0.0038, ..., -0.0051, -0.0101, 0.0044]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.k_norm.weight': tensor([ 1.5156e+00, 7.0703e-01, 3.8867e-01, 2.8711e-01, 1.3828e+00, + -7.5989e-03, 2.1484e-02, -4.3511e-06, 2.9663e-02, -1.7471e-03, + -1.0943e-07, -2.4959e-07, 1.1484e+00, 3.7537e-03, -3.7126e-13, + -1.6602e-02, -9.8419e-04, 8.0078e-02, 3.2227e-02, -3.7384e-03, + 5.3101e-03, 9.2030e-05, 5.8594e-01, 1.2344e+00, 8.6426e-02, + 5.4297e-01, 4.4141e-01, 1.6406e-01, 4.8828e-03, -5.6763e-03, + 1.4709e-02, -7.8201e-04, 1.3306e-02, 4.9561e-02, 1.9226e-03, + -1.7014e-03, 8.1787e-03, 5.9128e-04, 4.0894e-03, 4.0829e-06, + -1.1444e-04, 1.4114e-03, 2.2583e-03, 1.2891e+00, 6.6797e-01, + 1.3125e+00, 5.8203e-01, 2.9492e-01, -4.8828e-03, 6.2012e-02, + 3.7231e-03, -6.1951e-03, 3.3722e-03, -1.6928e-05, 6.5804e-05, + 1.1697e-06, -2.0885e-04, -3.1710e-05, -9.8877e-03, -3.7835e-09, + 5.4932e-03, -1.6975e-04, -3.5645e-02, -1.4099e-02, 7.0312e-01, + 1.3359e+00, 1.3359e+00, 1.3828e+00, 2.5977e-01, 1.5259e-02, + -2.2125e-03, 8.3447e-06, -2.8687e-03, -1.3184e-02, -1.3318e-07, + 2.3246e-06, 5.6763e-03, 3.0365e-03, 3.0376e-13, 8.1250e-01, + 1.0625e+00, 6.7871e-02, 1.3733e-04, -4.4861e-03, -4.0894e-03, + 9.9487e-03, 1.2500e+00, 7.3047e-01, 1.4766e+00, 1.5312e+00, + 4.5410e-02, 4.8828e-01, -1.4160e-02, 2.1118e-02, 1.6251e-03, + -6.7234e-05, 1.3794e-02, 5.3223e-02, 1.0071e-02, -5.9204e-03, + 8.5449e-03, -3.3617e-05, 1.1562e+00, 6.2212e-07, -8.0585e-05, + -3.8300e-03, 4.7386e-06, 3.7891e-01, 1.2891e+00, 1.3672e+00, + 1.6953e+00, 5.3125e-01, -2.5024e-02, 8.5938e-02, 6.5613e-03, + 1.7334e-02, -5.0964e-03, 1.1146e-05, 1.3709e-05, 5.3644e-07, + -2.5749e-05, -1.0529e-03, 1.1250e+00, -5.0664e-07, -7.2021e-03, + 2.2411e-05, 7.8125e-01, 1.1016e+00], dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.self_attn.v_proj.weight': tensor([[-0.0078, 0.0123, -0.0081, ..., -0.0261, 0.0009, 0.0117], + [ 0.0104, -0.0057, -0.0150, ..., 0.0039, -0.0039, 0.0118], + [-0.0068, -0.0043, 0.0041, ..., 0.0190, -0.0097, -0.0103], + ..., + [ 0.0154, -0.0128, -0.0153, ..., -0.0254, -0.0073, 0.0041], + [ 0.0048, -0.0039, 0.0100, ..., 0.0019, 0.0038, -0.0142], + [-0.0115, 0.0057, -0.0239, ..., 0.0041, 0.0121, 0.0092]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.self_attn.output_proj.weight': tensor([[ 0.0040, 0.0135, -0.0322, ..., 0.0049, -0.0205, 0.0128], + [ 0.0024, -0.0211, -0.0120, ..., 0.0063, 0.0154, -0.0089], + [-0.0041, 0.0023, 0.0042, ..., 0.0028, 0.0031, 0.0042], + ..., + [ 0.0118, 0.0081, -0.0229, ..., 0.0190, 0.0114, 0.0009], + [ 0.0064, 0.0043, 0.0041, ..., 0.0077, 0.0093, 0.0035], + [ 0.0062, 0.0047, -0.0041, ..., -0.0068, 0.0107, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_proj.weight': tensor([[-0.0069, 0.0082, 0.0011, ..., -0.0103, -0.0052, 0.0002], + [ 0.0091, 0.0072, 0.0038, ..., -0.0041, -0.0059, 0.0029], + [ 0.0051, 0.0062, -0.0024, ..., -0.0035, 0.0157, -0.0038], + ..., + [ 0.0027, 0.0070, 0.0029, ..., 0.0079, 0.0060, 0.0046], + [-0.0107, 0.0010, 0.0054, ..., 0.0030, -0.0099, 0.0085], + [ 0.0070, 0.0062, -0.0022, ..., -0.0051, 0.0134, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_norm.weight': tensor([3.5400e-03, 1.4551e-01, 1.3245e-02, 5.8594e-01, 2.1240e-02, 2.3315e-02, + 8.7891e-03, 4.8523e-03, 1.7456e-02, 1.2598e-01, 8.4839e-03, 5.5664e-02, + 2.7344e-01, 1.9165e-02, 1.8945e-01, 1.5320e-02, 6.7871e-02, 1.5442e-02, + 4.1504e-03, 1.8555e-02, 1.0559e-02, 1.8652e-01, 2.8711e-01, 1.9922e-01, + 2.3315e-02, 1.5918e-01, 6.6406e-02, 1.6357e-02, 4.0625e-01, 8.6670e-03, + 1.9775e-02, 5.1172e-01, 6.6895e-02, 3.4668e-02, 1.5991e-02, 1.5991e-02, + 7.5195e-02, 4.3359e-01, 5.3125e-01, 1.2451e-02, 8.6060e-03, 1.2109e-01, + 5.6152e-03, 5.2344e-01, 3.0273e-02, 5.0293e-02, 1.6235e-02, 3.8086e-01, + 7.6294e-03, 1.8555e-02, 8.2520e-02, 1.2451e-02, 1.8799e-02, 4.6484e-01, + 1.8768e-03, 2.6978e-02, 5.4932e-02, 8.0078e-02, 6.2256e-03, 1.5991e-02, + 4.8523e-03, 2.3926e-02, 1.0559e-02, 3.6133e-01, 1.6602e-02, 3.0670e-03, + 2.1118e-02, 5.9570e-02, 5.9814e-02, 1.3611e-02, 6.6528e-03, 6.0156e-01, + 3.7575e-04, 5.7031e-01, 1.3855e-02, 3.9673e-03, 1.6602e-02, 3.0664e-01, + 2.6733e-02, 3.3722e-03, 1.0986e-01, 5.2490e-03, 6.3477e-02, 9.1553e-03, + 1.9897e-02, 8.1787e-03, 1.2390e-02, 6.5430e-02, 2.0508e-01, 1.0315e-02, + 1.5564e-03, 1.8433e-02, 4.6387e-02, 2.4170e-02, 2.6245e-02, 1.5747e-02, + 1.7212e-02, 2.3193e-02, 6.3965e-02, 2.7344e-02, 4.3213e-02, 1.5198e-02, + 3.1738e-03, 6.5430e-02, 5.8594e-01, 7.6172e-02, 6.8359e-03, 4.8047e-01, + 1.3428e-02, 3.0762e-02, 1.7456e-02, 1.3000e-02, 1.4591e-04, 9.0332e-02, + 4.7852e-01, 2.4170e-02, 2.4292e-02, 5.7422e-01, 1.2158e-01, 2.6001e-02, + 1.9775e-02, 6.8750e-01, 1.5234e-01, 1.5869e-02, 1.4404e-02, 5.4932e-03, + 2.8442e-02, 1.1292e-02], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_proj.weight': tensor([[ 2.1057e-03, -3.3875e-03, 4.6692e-03, ..., 6.9427e-04, + -4.0283e-03, 4.7302e-03], + [-7.0190e-03, 2.6321e-04, -5.3711e-03, ..., 5.0964e-03, + -7.8125e-03, -5.4626e-03], + [-5.3711e-03, -1.6174e-03, -5.2490e-03, ..., 1.2451e-02, + -1.4526e-02, -3.6926e-03], + ..., + [ 6.9580e-03, 4.6082e-03, 2.5482e-03, ..., 2.9297e-03, + -1.8768e-03, -6.6223e-03], + [ 6.2866e-03, 2.9449e-03, 1.3504e-03, ..., -1.3275e-03, + 2.6398e-03, 4.4250e-03], + [ 6.9141e-05, 6.7902e-04, -1.2589e-03, ..., -6.3782e-03, + -6.1340e-03, 3.0823e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_norm.weight': tensor([ 3.5400e-03, 1.4551e-01, 1.3062e-02, 5.8594e-01, 2.0996e-02, + 2.3193e-02, 8.9722e-03, 4.7302e-03, 1.7700e-02, 1.2598e-01, + 8.4839e-03, 5.3955e-02, 2.7344e-01, 1.9287e-02, 1.8945e-01, + 1.6113e-02, 6.7383e-02, 1.5442e-02, 4.0894e-03, 1.8555e-02, + 1.0803e-02, 1.8750e-01, 2.8906e-01, 1.9922e-01, 2.3071e-02, + 1.5918e-01, 6.5430e-02, 1.5137e-02, 4.0820e-01, 8.6670e-03, + 2.0020e-02, 5.1172e-01, 6.6895e-02, 3.4424e-02, 1.6357e-02, + 1.5625e-02, 7.5195e-02, 4.3359e-01, 5.3125e-01, 1.2451e-02, + 8.5449e-03, 1.2109e-01, 5.5847e-03, 5.2344e-01, 3.0762e-02, + 5.0293e-02, 1.6479e-02, 3.8086e-01, 7.7209e-03, 1.8433e-02, + 8.2520e-02, 1.2817e-02, 1.8799e-02, 4.6289e-01, 2.1057e-03, + 2.6978e-02, 5.4688e-02, 8.0078e-02, 6.3171e-03, 1.4343e-02, + 4.7913e-03, 2.3804e-02, 1.0681e-02, 3.6133e-01, 1.6479e-02, + 3.1281e-03, 2.0630e-02, 5.9570e-02, 5.9814e-02, 1.3611e-02, + 7.0496e-03, 6.0156e-01, 1.6022e-04, 5.7031e-01, 1.4038e-02, + -4.2114e-03, 1.6602e-02, 3.0664e-01, 2.7344e-02, 3.4637e-03, + 1.1035e-01, 5.0964e-03, 6.2988e-02, 8.8501e-03, 1.9409e-02, + 8.2397e-03, 1.3000e-02, 6.5918e-02, 2.0508e-01, 1.0193e-02, + 1.7548e-03, 1.7212e-02, 4.6631e-02, 2.4170e-02, 2.6978e-02, + 1.5869e-02, 1.7334e-02, 2.3193e-02, 6.3965e-02, 2.7344e-02, + 4.2725e-02, 1.5320e-02, 3.8452e-03, 6.5918e-02, 5.8594e-01, + 7.6172e-02, 6.6223e-03, 4.8047e-01, 1.3062e-02, 3.0518e-02, + 1.8066e-02, 1.2939e-02, 1.5259e-04, 9.0332e-02, 4.7852e-01, + 2.3682e-02, 2.4292e-02, 5.7422e-01, 1.2305e-01, 2.6245e-02, + 1.9653e-02, 6.9141e-01, 1.5137e-01, 1.4648e-02, 1.3916e-02, + 5.4321e-03, 2.7954e-02, 1.1108e-02], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.v_proj.weight': tensor([[ 0.0136, -0.0051, -0.0161, ..., 0.0037, -0.0212, -0.0067], + [-0.0162, -0.0009, -0.0035, ..., -0.0041, -0.0212, 0.0166], + [-0.0165, -0.0024, 0.0033, ..., 0.0016, 0.0272, 0.0325], + ..., + [ 0.0039, -0.0062, -0.0161, ..., 0.0018, 0.0115, 0.0039], + [ 0.0231, 0.0065, 0.0222, ..., 0.0047, 0.0070, 0.0078], + [ 0.0183, 0.0043, 0.0031, ..., 0.0067, -0.0141, -0.0238]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.output_proj.weight': tensor([[ 0.0136, 0.0078, -0.0025, ..., -0.0070, 0.0131, 0.0087], + [-0.0044, -0.0047, 0.0051, ..., 0.0121, 0.0042, -0.0007], + [ 0.0028, -0.0032, 0.0048, ..., -0.0025, 0.0024, -0.0036], + ..., + [-0.0141, -0.0090, -0.0069, ..., 0.0048, -0.0078, 0.0093], + [ 0.0165, -0.0115, 0.0109, ..., -0.0265, -0.0118, -0.0236], + [ 0.0045, -0.0073, 0.0156, ..., 0.0239, 0.0071, 0.0122]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0066, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [ 0.0018, 0.0150, -0.0159, ..., 0.0040, -0.0058, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0070, -0.0072], + [-0.0337, -0.0037, 0.0052, ..., -0.0039, 0.0172, 0.0231], + [-0.0255, -0.0015, -0.0039, ..., 0.0408, 0.0430, -0.0123]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0031], + [ 0.0198, -0.0208, 0.0129, ..., 0.0088, -0.0337, -0.0051], + [ 0.0302, 0.0101, 0.0025, ..., -0.0304, 0.0199, -0.0019], + ..., + [ 0.0171, -0.0276, -0.0075, ..., 0.0231, 0.0182, -0.0045], + [ 0.0221, 0.0388, 0.0060, ..., 0.0037, -0.0488, -0.0361], + [ 0.0040, 0.0115, -0.0058, ..., 0.0037, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img.weight': tensor([[ 0.0273, 0.0038, 0.0118, ..., -0.0015, 0.0347, 0.0615], + [-0.0258, 0.0009, -0.0197, ..., 0.0139, -0.0045, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0096, 0.0132, 0.0184, ..., 0.0110, 0.0084, 0.0515], + [-0.0010, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.2.mlp.layer1.weight': tensor([[-0.0015, 0.0028, 0.0081, ..., 0.0044, -0.0105, -0.0049], + [-0.0078, 0.0064, 0.0078, ..., -0.0059, -0.0128, 0.0036], + [-0.0074, 0.0156, 0.0066, ..., 0.0040, -0.0019, -0.0508], + ..., + [-0.0068, 0.0077, -0.0031, ..., 0.0070, 0.0021, -0.0077], + [ 0.0013, -0.0021, -0.0081, ..., -0.0139, -0.0040, -0.0269], + [ 0.0014, 0.0190, 0.0085, ..., 0.0118, 0.0143, -0.0080]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.mlp.layer2.weight': tensor([[-0.0039, -0.0052, -0.0156, ..., -0.0110, -0.0114, 0.0038], + [ 0.0078, -0.0022, 0.0026, ..., -0.0114, -0.0106, 0.0042], + [-0.0045, 0.0028, -0.0003, ..., 0.0067, 0.0045, 0.0061], + ..., + [-0.0017, -0.0100, -0.0094, ..., 0.0044, 0.0038, 0.0044], + [-0.0137, -0.0057, 0.0025, ..., 0.0140, 0.0033, -0.0125], + [ 0.0079, 0.0063, -0.0026, ..., -0.0060, 0.0133, -0.0347]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0299, 0.0299, 0.0286, ..., -0.0041, -0.0042, -0.0043], + [-0.0284, -0.0275, -0.0271, ..., 0.0042, 0.0046, 0.0045], + [-0.0298, -0.0292, -0.0287, ..., 0.0044, 0.0048, 0.0048], + ..., + [-0.0344, -0.0332, -0.0327, ..., 0.0020, 0.0021, 0.0023], + [-0.0337, -0.0332, -0.0325, ..., 0.0020, 0.0021, 0.0020], + [-0.0156, -0.0157, -0.0159, ..., 0.0001, 0.0001, 0.0001]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0055, -0.0037, -0.0053, ..., -0.0017, -0.0019, -0.0014], + [ 0.0044, -0.0040, -0.0042, ..., -0.0033, -0.0038, -0.0022], + [ 0.0057, -0.0066, -0.0056, ..., -0.0018, -0.0021, 0.0015], + ..., + [ 0.0420, -0.0452, -0.0454, ..., -0.0278, -0.0244, 0.0075], + [-0.0060, 0.0142, 0.0075, ..., 0.0167, 0.0153, 0.0170], + [-0.0189, 0.0197, 0.0195, ..., 0.0182, 0.0177, 0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5420e-02, 5.4443e-02, 5.3467e-02, ..., -1.3733e-04, + -6.7234e-05, -2.1935e-04], + [ 5.4688e-02, 5.2979e-02, 5.1025e-02, ..., 2.7275e-04, + 3.8528e-04, 4.1771e-04], + [-4.5654e-02, -4.4434e-02, -4.3213e-02, ..., -3.8445e-06, + 5.1498e-05, -5.6744e-05], + ..., + [ 2.8809e-02, 2.7954e-02, 2.7466e-02, ..., 1.0109e-04, + 3.3140e-05, 6.2943e-05], + [ 5.9814e-02, 5.8594e-02, 5.7861e-02, ..., 4.6921e-04, + 5.1117e-04, 5.0354e-04], + [-4.7607e-02, -4.6631e-02, -4.5654e-02, ..., -2.9755e-04, + -3.4142e-04, -3.0327e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0041, 0.0135, -0.0018, ..., -0.0096, -0.0148, 0.0106], + [-0.0134, 0.0061, 0.0108, ..., -0.0055, -0.0068, 0.0040], + [-0.0006, 0.0032, 0.0011, ..., -0.0012, -0.0017, 0.0017], + ..., + [ 0.0178, 0.0061, -0.0118, ..., 0.0077, 0.0155, 0.0041], + [-0.0043, 0.0393, 0.0046, ..., -0.0013, 0.0275, -0.0136], + [-0.0101, 0.0148, 0.0110, ..., -0.0106, -0.0070, 0.0193]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_mlp.1.weight': tensor([[-0.0518, -0.0491, -0.0476, ..., 0.0030, 0.0032, 0.0032], + [-0.0063, 0.0039, 0.0039, ..., 0.0007, 0.0008, 0.0007], + [ 0.0645, 0.0625, 0.0623, ..., -0.0016, -0.0016, -0.0016], + ..., + [ 0.0588, 0.0574, 0.0566, ..., -0.0017, -0.0017, -0.0015], + [ 0.0151, 0.0139, 0.0133, ..., -0.0010, -0.0009, -0.0010], + [-0.0598, -0.0586, -0.0576, ..., 0.0021, 0.0021, 0.0020]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.adaln_modulation_mlp.2.weight': tensor([[-0.0337, -0.0108, 0.0143, ..., 0.0069, 0.0125, -0.0147], + [ 0.0018, -0.0047, -0.0014, ..., -0.0011, 0.0047, 0.0011], + [ 0.0029, -0.0035, -0.0018, ..., -0.0018, 0.0041, 0.0018], + ..., + [ 0.0403, -0.0041, -0.0137, ..., -0.0083, 0.0041, 0.0143], + [ 0.0270, 0.0325, 0.0256, ..., 0.0282, -0.0325, -0.0258], + [ 0.0095, -0.0148, -0.0086, ..., -0.0112, 0.0151, 0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.after_proj.weight': tensor([[-0.0071, -0.0042, 0.0067, ..., -0.0049, 0.0068, -0.0044], + [ 0.0003, 0.0037, 0.0046, ..., -0.0112, -0.0070, -0.0019], + [-0.0046, -0.0081, 0.0173, ..., 0.0039, 0.0036, 0.0035], + ..., + [-0.0026, 0.0045, 0.0090, ..., -0.0181, 0.0084, 0.0073], + [ 0.0036, -0.0075, -0.0081, ..., 0.0049, -0.0157, 0.0065], + [-0.0063, -0.0060, 0.0150, ..., -0.0077, 0.0002, 0.0079]], + dtype=torch.bfloat16), 'net.control_blocks_1.2.after_proj.bias': tensor([-0.0017, -0.0015, 0.0014, ..., -0.0042, 0.0008, 0.0029], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_proj.weight': tensor([[-0.0034, 0.0080, 0.0164, ..., -0.0070, 0.0063, -0.0270], + [-0.0066, 0.0023, -0.0067, ..., -0.0078, -0.0089, -0.0082], + [ 0.0114, 0.0062, 0.0042, ..., -0.0023, -0.0038, -0.0038], + ..., + [ 0.0045, -0.0076, -0.0040, ..., -0.0042, 0.0046, 0.0069], + [-0.0044, -0.0320, -0.0098, ..., -0.0381, -0.0189, -0.0298], + [-0.0075, -0.0038, -0.0040, ..., 0.0037, -0.0047, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_norm.weight': tensor([ 1.2656e+00, 1.6562e+00, 1.5078e+00, 1.7188e+00, 1.0498e-01, + 6.9824e-02, 4.7607e-02, 1.0192e-05, -1.2757e-30, -1.5564e-03, + 2.6978e-02, 1.0156e-01, -2.2095e-02, -2.9297e-02, 9.0790e-04, + 3.2616e-04, 6.0059e-02, 9.4922e-01, 9.9609e-01, 1.4267e-03, + 6.1035e-05, 1.1953e+00, 1.6406e+00, 1.6719e+00, 6.7969e-01, + 6.2891e-01, 1.5703e+00, -3.4332e-03, -1.0071e-02, 2.0146e-05, + 4.3869e-04, -5.0354e-03, 2.3682e-02, 1.0938e+00, 5.4932e-03, + 1.0781e+00, 1.5068e-04, 4.7207e-05, 1.3123e-02, -1.6327e-03, + 8.5449e-03, -3.2187e-05, 1.2054e-03, 1.6953e+00, 9.8438e-01, + 1.4062e+00, 1.6406e+00, 1.8516e+00, 1.0596e-01, 4.9973e-04, + 3.0029e-02, 2.2168e-01, 1.0859e+00, 8.2520e-02, 2.1820e-03, + 1.3086e-01, 5.6458e-04, 9.3750e-02, 7.4863e-05, -1.1902e-02, + 1.4267e-03, 3.8300e-03, -3.3722e-03, 1.1406e+00, 1.8828e+00, + 7.9297e-01, 9.5703e-01, 1.5312e+00, 4.0771e-02, -1.2390e-02, + 1.2085e-02, -1.9908e-05, 2.5761e-30, 1.1658e-02, 1.7822e-02, + 8.9722e-03, 6.0730e-03, 4.0588e-03, -2.5269e-02, 1.8066e-02, + 5.7861e-02, -2.6245e-03, 1.4587e-02, 2.7618e-03, 3.3379e-04, + 4.1504e-03, 5.5469e-01, 4.2578e-01, 1.4688e+00, 1.6406e+00, + 4.7656e-01, 5.7068e-03, -7.9727e-04, 2.9755e-04, -4.3335e-03, + 1.0300e-03, 3.4912e-02, 6.8848e-02, 5.2795e-03, 5.2002e-02, + 2.6321e-04, 4.1485e-05, 1.2891e-01, -2.6321e-04, 6.5918e-03, + -1.4484e-05, 1.3638e-04, 4.1797e-01, 1.5000e+00, 1.4062e+00, + 5.9375e-01, 9.2969e-01, 7.3242e-02, -2.4109e-03, 2.0020e-02, + 1.1641e+00, 5.0049e-02, 1.0859e+00, -1.5736e-05, 2.3651e-03, + 3.9307e-02, 4.5410e-02, 1.4210e-04, 1.9302e-03, -1.0872e-04, + -3.6011e-03, 9.4141e-01, -2.0264e-02], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.self_attn.k_proj.weight': tensor([[-4.2725e-02, 1.7700e-03, -2.9449e-03, ..., -3.2959e-03, + -3.9101e-05, -1.3123e-02], + [-3.5156e-02, -1.2817e-02, -1.3123e-02, ..., 1.0071e-02, + -1.1230e-02, 2.4414e-02], + [ 4.5410e-02, 4.3030e-03, 3.8910e-03, ..., -1.2146e-02, + -4.1809e-03, -2.7100e-02], + ..., + [-7.1411e-03, 1.5259e-02, 5.7602e-04, ..., -5.2795e-03, + -6.2561e-03, -2.8076e-03], + [ 1.1597e-02, -8.8501e-03, -2.4261e-03, ..., -4.0894e-03, + 1.0986e-02, 3.2349e-03], + [ 2.0294e-03, 5.9814e-03, -1.2970e-03, ..., -5.9509e-03, + -6.1035e-04, 6.5308e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.k_norm.weight': tensor([ 7.4609e-01, 1.5938e+00, 1.5234e+00, 1.6562e+00, 1.0742e-01, + 7.5195e-02, 4.6387e-02, -1.3113e-05, 8.1351e-31, -2.1210e-03, + 2.6611e-02, 1.0205e-01, -2.6245e-02, 5.2979e-02, 1.1597e-03, + 1.4114e-04, 4.0588e-03, 9.4922e-01, 9.9609e-01, 1.3275e-03, + 1.0681e-04, 1.1953e+00, 1.4531e+00, 1.5469e+00, 3.5400e-02, + 1.4355e-01, 1.5938e+00, -5.7373e-03, 9.6436e-03, -9.6798e-05, + -1.5640e-03, 5.0964e-03, -2.0630e-02, 1.0938e+00, -2.0142e-03, + 1.0703e+00, 1.6975e-04, 4.4584e-05, -1.5076e-02, -1.7395e-03, + 8.5449e-03, -7.4387e-05, 1.2054e-03, 1.4609e+00, 4.4727e-01, + 1.3750e+00, 1.6953e+00, 1.8672e+00, -5.3955e-02, 1.8387e-03, + 1.0681e-02, 1.5625e-02, 1.0859e+00, 7.6172e-02, 2.1820e-03, + 1.3086e-01, -9.0122e-05, 9.3750e-02, 7.0095e-05, 7.6660e-02, + -1.4191e-03, 1.8311e-02, -1.9653e-02, 1.1406e+00, 1.9688e+00, + 1.1172e+00, 8.6328e-01, 1.5781e+00, 2.7710e-02, -5.7983e-03, + -1.3916e-02, 1.5974e-05, 1.9352e-30, 1.2024e-02, 3.1006e-02, + 5.0781e-02, 6.3782e-03, 2.6611e-02, -3.7842e-02, -1.8066e-02, + -1.8555e-02, 6.0730e-03, 9.0332e-03, 2.6550e-03, 3.2425e-04, + 8.6670e-03, 1.1172e+00, 1.0000e+00, 1.5625e+00, 1.7031e+00, + 1.4844e-01, 5.7678e-03, -3.0212e-03, -2.6131e-04, -3.6163e-03, + -6.2943e-04, 1.9287e-02, 9.0332e-03, 4.6387e-03, -8.9722e-03, + 2.7084e-04, -1.6332e-05, 1.2793e-01, -2.6321e-04, 6.5613e-03, + 2.1100e-05, 1.3256e-04, 1.1719e+00, 1.6406e+00, 1.4141e+00, + 1.1768e-01, 9.8438e-01, 8.7280e-03, 4.0894e-03, 5.0354e-03, + 1.1641e+00, -2.9755e-03, 1.0938e+00, -7.2002e-05, -1.8158e-03, + 3.9307e-02, 4.5166e-02, 1.3924e-04, 4.3457e-02, -1.1730e-04, + 1.5015e-02, 9.4141e-01, -1.5503e-02], dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.self_attn.v_proj.weight': tensor([[-0.0017, 0.0042, 0.0115, ..., 0.0542, -0.0092, 0.0073], + [ 0.0060, 0.0214, -0.0223, ..., -0.0248, -0.0010, -0.0054], + [-0.0009, 0.0156, -0.0039, ..., 0.0126, -0.0044, 0.0096], + ..., + [ 0.0026, -0.0039, -0.0148, ..., 0.0102, 0.0055, -0.0009], + [ 0.0059, 0.0243, 0.0019, ..., 0.0025, 0.0025, 0.0002], + [-0.0074, -0.0142, 0.0027, ..., 0.0175, -0.0077, -0.0128]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.self_attn.output_proj.weight': tensor([[ 0.0179, 0.0205, -0.0031, ..., 0.0021, 0.0260, -0.0159], + [ 0.0096, -0.0099, -0.0095, ..., -0.0045, -0.0166, 0.0025], + [ 0.0078, 0.0236, -0.0091, ..., 0.0062, -0.0017, -0.0050], + ..., + [-0.0287, -0.0366, -0.0045, ..., -0.0045, -0.0148, 0.0040], + [ 0.0026, 0.0087, -0.0182, ..., -0.0103, 0.0327, -0.0085], + [-0.0432, -0.0049, 0.0156, ..., 0.0238, -0.0083, -0.0036]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_proj.weight': tensor([[ 0.0049, 0.0067, 0.0038, ..., -0.0072, -0.0022, 0.0223], + [ 0.0061, -0.0040, -0.0039, ..., -0.0063, 0.0070, 0.0106], + [ 0.0023, 0.0029, -0.0030, ..., 0.0027, 0.0157, 0.0051], + ..., + [ 0.0076, -0.0088, -0.0059, ..., 0.0037, -0.0072, -0.0011], + [ 0.0017, -0.0032, -0.0029, ..., 0.0056, -0.0051, -0.0022], + [-0.0040, -0.0054, 0.0039, ..., -0.0080, 0.0072, 0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_norm.weight': tensor([0.1089, 0.0232, 0.4688, 0.0752, 0.1816, 0.0356, 0.1494, 0.1260, 0.5234, + 0.0388, 0.0630, 0.0148, 0.0630, 0.0762, 0.0508, 0.2021, 0.4883, 0.0260, + 0.7383, 0.0410, 0.0311, 0.1641, 0.0289, 0.0806, 0.0557, 0.4199, 0.5156, + 0.2754, 0.0713, 0.0454, 0.0233, 0.2451, 0.7031, 0.2617, 0.0415, 0.0344, + 0.0190, 0.1025, 0.0284, 0.0422, 0.0586, 0.1318, 0.0247, 0.0344, 0.2246, + 0.0425, 0.4043, 0.0304, 0.0540, 0.0095, 0.0259, 0.4141, 0.4902, 0.3867, + 0.0962, 0.6133, 0.5781, 0.0337, 0.1494, 0.0249, 0.0898, 0.4902, 0.0215, + 0.5547, 0.1699, 0.4180, 0.1040, 0.0488, 0.0723, 0.4941, 0.0262, 0.4902, + 0.0198, 0.0447, 0.1436, 0.0850, 0.0280, 0.1128, 0.6445, 0.0308, 0.0908, + 0.0874, 0.0417, 0.0381, 0.4297, 0.0723, 0.0515, 0.0874, 0.0211, 0.1787, + 0.0398, 0.1924, 0.0403, 0.0669, 0.0442, 0.1133, 0.0052, 0.3691, 0.5234, + 0.0471, 0.0581, 0.3301, 0.0347, 0.1270, 0.0352, 0.2656, 0.0214, 0.5664, + 0.5664, 0.0391, 0.0176, 0.7500, 0.0310, 0.0139, 0.5039, 0.2217, 0.3145, + 0.7109, 0.0332, 0.0913, 0.0884, 0.5391, 0.6016, 0.0146, 0.0248, 0.2041, + 0.0415, 0.0378], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_proj.weight': tensor([[-1.7548e-03, 1.8158e-03, 3.5706e-03, ..., -3.2349e-03, + 1.1353e-02, -1.2512e-02], + [ 5.9204e-03, -6.7711e-05, -2.9602e-03, ..., -3.0518e-04, + -1.3275e-03, 6.3171e-03], + [-1.1292e-02, 7.6904e-03, -1.3245e-02, ..., 6.8970e-03, + 9.8267e-03, 1.7853e-03], + ..., + [-1.6479e-03, 3.1128e-03, -5.8899e-03, ..., -3.5095e-03, + -1.3306e-02, 9.3384e-03], + [-1.2451e-02, -1.8539e-03, -8.9111e-03, ..., 7.5684e-03, + -9.2163e-03, -2.7275e-04], + [ 1.4893e-02, -1.4038e-03, 5.0049e-03, ..., -1.0071e-02, + 4.8523e-03, -3.5400e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_norm.weight': tensor([0.1084, 0.0232, 0.4688, 0.0752, 0.1816, 0.0361, 0.1494, 0.1260, 0.5234, + 0.0393, 0.0640, 0.0149, 0.0630, 0.0762, 0.0510, 0.2041, 0.4883, 0.0262, + 0.7344, 0.0410, 0.0309, 0.1641, 0.0291, 0.0801, 0.0557, 0.4160, 0.5156, + 0.2734, 0.0713, 0.0452, 0.0242, 0.2471, 0.7031, 0.2617, 0.0422, 0.0344, + 0.0188, 0.1025, 0.0287, 0.0417, 0.0583, 0.1328, 0.0251, 0.0349, 0.2246, + 0.0425, 0.4043, 0.0306, 0.0542, 0.0092, 0.0262, 0.4141, 0.4902, 0.3867, + 0.0962, 0.6133, 0.5820, 0.0334, 0.1494, 0.0253, 0.0898, 0.4902, 0.0219, + 0.5547, 0.1689, 0.4180, 0.1040, 0.0488, 0.0723, 0.4961, 0.0261, 0.4902, + 0.0195, 0.0444, 0.1436, 0.0845, 0.0280, 0.1138, 0.6445, 0.0305, 0.0898, + 0.0874, 0.0417, 0.0383, 0.4297, 0.0728, 0.0505, 0.0869, 0.0214, 0.1807, + 0.0408, 0.1914, 0.0400, 0.0674, 0.0447, 0.1143, 0.0052, 0.3691, 0.5234, + 0.0464, 0.0588, 0.3301, 0.0347, 0.1270, 0.0352, 0.2656, 0.0211, 0.5664, + 0.5664, 0.0391, 0.0177, 0.7500, 0.0305, 0.0141, 0.5078, 0.2217, 0.3145, + 0.7070, 0.0325, 0.0898, 0.0884, 0.5391, 0.6016, 0.0145, 0.0248, 0.2041, + 0.0427, 0.0381], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.v_proj.weight': tensor([[ 0.0099, -0.0024, -0.0236, ..., -0.0018, -0.0025, -0.0165], + [-0.0064, 0.0107, 0.0040, ..., 0.0040, 0.0023, -0.0250], + [ 0.0190, 0.0034, -0.0056, ..., 0.0033, -0.0052, -0.0396], + ..., + [ 0.0025, 0.0031, 0.0190, ..., -0.0087, -0.0259, 0.0049], + [-0.0332, -0.0007, -0.0219, ..., -0.0072, 0.0103, -0.0148], + [-0.0070, -0.0017, -0.0017, ..., -0.0082, 0.0104, 0.0071]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.output_proj.weight': tensor([[-0.0128, -0.0076, -0.0082, ..., -0.0063, 0.0133, 0.0111], + [-0.0248, -0.0042, -0.0100, ..., 0.0048, -0.0138, -0.0304], + [ 0.0011, 0.0024, 0.0064, ..., 0.0029, -0.0026, -0.0045], + ..., + [ 0.0141, -0.0109, 0.0123, ..., 0.0160, -0.0127, -0.0093], + [-0.0161, -0.0079, 0.0060, ..., 0.0239, -0.0284, 0.0334], + [-0.0165, -0.0115, 0.0254, ..., -0.0051, 0.0259, -0.0051]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_img.weight': tensor([[ 1.6235e-02, -4.2480e-02, -2.7954e-02, ..., 2.3682e-02, + -4.1748e-02, -4.6631e-02], + [-1.0925e-02, 2.9297e-02, 2.5635e-02, ..., -2.3315e-02, + -1.9775e-02, -2.3193e-02], + [-1.2146e-02, 4.9438e-03, -6.5918e-03, ..., 2.6611e-02, + -2.7832e-02, 7.5989e-03], + ..., + [-2.1729e-02, -4.1016e-02, 4.8828e-02, ..., 3.8862e-05, + -2.3956e-03, -2.5513e-02], + [ 9.9182e-04, 6.6833e-03, 3.8300e-03, ..., 3.0029e-02, + 8.4229e-03, 2.0508e-02], + [-2.1362e-02, 2.1729e-02, 5.6458e-03, ..., -1.3916e-02, + -1.4282e-02, -4.1992e-02]], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, 0.0013, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0024, 0.0293, ..., 0.0067, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0139, 0.0126, -0.0092], + ..., + [-0.0032, 0.0320, -0.0233, ..., -0.0062, -0.0028, 0.0188], + [ 0.0135, -0.0352, 0.0267, ..., -0.0305, -0.0039, 0.0018], + [-0.0205, 0.0140, 0.0039, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, 0.0005, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0056, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0015], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, 0.0013, 0.0012], + [-0.0141, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0023, ..., -0.0103, 0.0071, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_1.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_1.3.mlp.layer1.weight': tensor([[ 0.0104, 0.0049, -0.0160, ..., 0.0096, 0.0089, -0.0065], + [-0.0109, -0.0065, 0.0049, ..., -0.0120, -0.0062, -0.0079], + [-0.0034, 0.0359, -0.0165, ..., 0.0284, 0.0075, 0.0089], + ..., + [-0.0074, 0.0103, 0.0645, ..., -0.0041, 0.0085, -0.0117], + [-0.0038, 0.0049, 0.0042, ..., -0.0151, 0.0087, -0.0129], + [ 0.0055, 0.0066, -0.0072, ..., -0.0146, -0.0156, -0.0186]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.mlp.layer2.weight': tensor([[ 0.0111, -0.0063, 0.0005, ..., -0.0266, -0.0067, 0.0039], + [-0.0112, 0.0051, -0.0106, ..., -0.0027, 0.0101, -0.0044], + [ 0.0006, -0.0107, 0.0055, ..., 0.0058, 0.0098, -0.0042], + ..., + [ 0.0040, 0.0040, 0.0217, ..., -0.0044, 0.0064, -0.0055], + [ 0.0053, 0.0029, 0.0039, ..., 0.0110, 0.0036, 0.0095], + [-0.0168, -0.0084, 0.0029, ..., 0.0116, 0.0139, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0413, -0.0408, -0.0405, ..., 0.0009, 0.0007, 0.0008], + [-0.0369, -0.0366, -0.0356, ..., 0.0026, 0.0027, 0.0027], + [-0.0334, -0.0332, -0.0320, ..., 0.0021, 0.0022, 0.0022], + ..., + [ 0.0244, 0.0234, 0.0225, ..., -0.0037, -0.0038, -0.0037], + [ 0.0134, 0.0123, 0.0115, ..., -0.0032, -0.0031, -0.0031], + [-0.0344, -0.0337, -0.0330, ..., 0.0022, 0.0023, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0031, 0.0044, ..., 0.0063, 0.0090, 0.0045], + [-0.0028, -0.0012, -0.0015, ..., -0.0019, -0.0014, -0.0011], + [ 0.0024, -0.0069, -0.0049, ..., 0.0032, 0.0033, -0.0058], + ..., + [-0.0237, -0.0811, -0.0635, ..., 0.0327, 0.0332, -0.0723], + [-0.0161, -0.0140, -0.0139, ..., -0.0115, -0.0139, -0.0131], + [ 0.0098, 0.0067, 0.0069, ..., 0.0092, 0.0121, 0.0064]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_cross_attn.1.weight': tensor([[-7.9102e-02, -7.6660e-02, -7.5195e-02, ..., -9.5844e-05, + -1.1110e-04, -1.0157e-04], + [ 3.0273e-02, 2.9785e-02, 2.9297e-02, ..., 2.5481e-06, + 2.8610e-05, 1.4877e-04], + [ 2.5391e-02, 2.4658e-02, 2.3682e-02, ..., 2.4796e-04, + 2.7084e-04, 1.7738e-04], + ..., + [ 5.1514e-02, 5.0293e-02, 4.9805e-02, ..., -1.8692e-04, + -1.8120e-04, -1.1349e-04], + [ 5.5176e-02, 5.3711e-02, 5.2246e-02, ..., -1.8311e-04, + -1.2207e-04, -1.7071e-04], + [-1.6235e-02, -1.5747e-02, -1.5503e-02, ..., -5.0306e-05, + 1.0109e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0053, -0.0033, ..., 0.0281, 0.0317, 0.0069], + [ 0.0040, -0.0044, 0.0056, ..., -0.0009, 0.0025, -0.0019], + [ 0.0025, -0.0038, 0.0039, ..., -0.0026, -0.0026, -0.0044], + ..., + [-0.0069, -0.0052, 0.0403, ..., 0.0016, 0.0123, -0.0041], + [-0.0302, 0.0184, 0.0298, ..., 0.0115, 0.0130, -0.0046], + [ 0.0261, -0.0206, 0.0040, ..., 0.0048, 0.0107, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_mlp.1.weight': tensor([[-7.6172e-02, -7.3730e-02, -7.2754e-02, ..., 8.7261e-05, + 2.0409e-04, 1.5354e-04], + [ 5.8105e-02, 5.5420e-02, 5.5176e-02, ..., 3.8719e-04, + 4.6921e-04, 4.5776e-04], + [-9.2773e-02, -9.1309e-02, -8.9844e-02, ..., -4.7874e-04, + -4.1199e-04, -4.4250e-04], + ..., + [-7.4707e-02, -7.2754e-02, -7.1777e-02, ..., 1.1444e-04, + 2.1648e-04, 1.1492e-04], + [-6.0791e-02, -5.9326e-02, -5.8594e-02, ..., 1.5736e-04, + 1.8477e-05, 2.9922e-05], + [ 7.2266e-02, 7.1289e-02, 7.0312e-02, ..., -9.0599e-05, + -1.1873e-04, -8.2016e-05]], dtype=torch.bfloat16), 'net.control_blocks_1.3.adaln_modulation_mlp.2.weight': tensor([[-4.9438e-03, -3.9978e-03, 2.5177e-03, ..., -7.4158e-03, + -1.1414e-02, 8.9111e-03], + [-3.3417e-03, -2.6550e-03, -1.6861e-03, ..., -2.7618e-03, + -2.5787e-03, 2.5787e-03], + [-6.4850e-05, -2.4414e-03, -1.7471e-03, ..., -3.5477e-04, + -1.5259e-04, 1.3161e-04], + ..., + [-4.6143e-02, -6.9275e-03, -5.6396e-02, ..., -3.4424e-02, + -1.2695e-02, 2.6489e-02], + [ 8.1177e-03, -2.7344e-02, 3.1738e-02, ..., 6.6528e-03, + 5.4016e-03, -6.5613e-03], + [ 9.4604e-03, 7.5989e-03, 4.9438e-03, ..., 8.5449e-03, + 4.1199e-03, -6.7444e-03]], dtype=torch.bfloat16), 'net.control_blocks_1.3.after_proj.weight': tensor([[ 0.0049, 0.0079, -0.0066, ..., 0.0017, 0.0227, 0.0034], + [-0.0020, -0.0085, -0.0080, ..., 0.0134, -0.0012, -0.0102], + [-0.0118, 0.0030, -0.0039, ..., -0.0023, -0.0021, 0.0011], + ..., + [-0.0079, -0.0026, 0.0031, ..., -0.0070, 0.0021, -0.0082], + [-0.0040, 0.0076, 0.0044, ..., 0.0051, -0.0108, -0.0024], + [ 0.0119, 0.0110, 0.0015, ..., -0.0018, 0.0058, 0.0067]], + dtype=torch.bfloat16), 'net.control_blocks_1.3.after_proj.bias': tensor([ 0.0050, -0.0118, -0.0096, ..., -0.0070, 0.0018, 0.0004], + dtype=torch.bfloat16)} +[06-26 07:20:34|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:34|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_1.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:35|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 07:20:35|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 1 checkpoint. +[06-26 07:20:35|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/depth(626e6618-bfcd-4d9a-a077-1409e2ce353f) +[06-26 07:20:35|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 2 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/dea7737ca29dd8d9086413c6dc5724b8250a0bb4/general/depth/626e6618-bfcd-4d9a-a077-1409e2ce353f_ema_bf16.pt +[06-26 07:20:46|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0144, -0.0021, 0.0064], + [ 0.0036, 0.0057, -0.0327, ..., -0.0082, -0.0024, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0137, -0.0029, 0.0283], + ..., + [ 0.0010, 0.0125, 0.0248, ..., 0.0437, -0.0074, -0.0192], + [ 0.0277, 0.0146, 0.0036, ..., -0.0031, -0.0092, 0.0146], + [ 0.0437, 0.0366, 0.0085, ..., 0.0013, -0.0128, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0028, -0.0310, 0.0300, ..., -0.0071, -0.0042, -0.0339], + [-0.0103, -0.0048, 0.0115, ..., 0.0049, 0.0092, 0.0520], + [-0.0064, 0.0132, 0.0398, ..., -0.0097, -0.0179, -0.0070], + ..., + [-0.0046, 0.0175, -0.0094, ..., 0.0225, 0.0047, -0.0020], + [-0.0209, -0.0081, 0.0087, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0354, -0.0082, ..., 0.0123, 0.0356, -0.0186], + [ 0.0513, 0.0141, 0.0016, ..., 0.0376, 0.0532, 0.0011], + [ 0.0175, -0.0347, 0.0015, ..., -0.0064, 0.0047, -0.0283], + ..., + [ 0.0354, -0.0047, -0.0209, ..., -0.0156, 0.0262, -0.0114], + [-0.0015, 0.0244, 0.0356, ..., 0.0018, 0.0170, 0.0112], + [ 0.0486, 0.0175, 0.0042, ..., 0.0046, -0.0021, 0.0008]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0118, ..., -0.0075, -0.0337, 0.0079], + [ 0.0491, 0.0413, 0.0251, ..., -0.0041, 0.0393, 0.0035], + [-0.0162, -0.0005, 0.0134, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0014, 0.0261, 0.0042, ..., -0.0038, 0.0347, 0.0036], + [-0.0249, 0.0067, 0.0024, ..., 0.0096, 0.0388, -0.0205], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0123, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0048, -0.0270, -0.0166, ..., 0.0118, -0.0298, 0.0532], + [ 0.0034, 0.0036, -0.0126, ..., -0.0161, 0.0152, 0.0308], + [ 0.0145, 0.0086, 0.0137, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0067, -0.0630, 0.0027, ..., 0.0072, -0.0183, 0.0217], + [ 0.0039, 0.0032, 0.0258, ..., -0.0486, -0.0349, -0.0039], + [ 0.0182, -0.0081, 0.0197, ..., 0.0151, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0053, -0.0042, -0.0005, ..., -0.0007, -0.0017, -0.0121], + [-0.0152, 0.0251, 0.0432, ..., -0.0051, 0.0039, -0.0038], + [-0.0149, -0.0203, -0.0245, ..., -0.0118, -0.0325, 0.0029], + ..., + [ 0.0043, 0.0008, 0.0417, ..., 0.0280, 0.0022, -0.0024], + [-0.0039, 0.0037, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0231, -0.0364, -0.0220, ..., 0.0067, -0.0034, -0.0085]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0155, -0.0063, -0.0481, ..., 0.0449, 0.0212, 0.0128], + [ 0.0156, -0.0227, 0.0220, ..., -0.0219, 0.0005, 0.0081], + [-0.0156, -0.0098, 0.0430, ..., -0.0101, 0.0327, -0.0110], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0136], + [ 0.0021, 0.0166, -0.0061, ..., 0.0013, 0.0079, -0.0052], + [-0.0110, -0.0041, 0.0117, ..., 0.0047, -0.0034, -0.0077]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0035, -0.0386, -0.0120, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0075, 0.0286, ..., -0.0339, -0.0160, 0.0170], + [-0.0127, -0.0053, -0.0128, ..., 0.0214, 0.0111, -0.0044], + ..., + [-0.0153, -0.0172, -0.0137, ..., 0.0098, 0.0098, 0.0184], + [-0.0154, -0.0356, 0.0201, ..., -0.0069, -0.0117, 0.0332], + [-0.0396, -0.0121, -0.0090, ..., -0.0284, -0.0141, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0280, 0.0312, 0.0133, ..., -0.0195, 0.0114, -0.0153], + [ 0.0192, 0.0092, -0.0073, ..., 0.0022, -0.0127, -0.0198], + [-0.0249, 0.0109, -0.0137, ..., 0.0090, 0.0172, 0.0067], + ..., + [ 0.0466, -0.0045, 0.0007, ..., 0.0010, 0.0239, 0.0042], + [ 0.0547, -0.0194, -0.0108, ..., -0.0069, 0.0311, -0.0044], + [ 0.0051, 0.0104, -0.0095, ..., 0.0203, 0.0170, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0044, -0.0046, 0.0283, ..., -0.0109, -0.0045, -0.0182], + [ 0.0052, 0.0126, -0.0315, ..., -0.0457, -0.0069, 0.0109], + [ 0.0006, 0.0261, 0.0245, ..., 0.0046, 0.0084, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0155, 0.0058, -0.0077], + [ 0.0059, -0.0371, 0.0356, ..., -0.0082, 0.0410, -0.0178], + [ 0.0178, -0.0155, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0087, 0.0208, 0.0272, ..., 0.0137, 0.0156, 0.0082], + [-0.0082, -0.0023, -0.0041, ..., 0.0144, 0.0007, 0.0057], + [ 0.0046, 0.0010, 0.0240, ..., -0.0119, -0.0066, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0046, ..., 0.0248, -0.0106, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., 0.0020, -0.0295, 0.0400], + [-0.0118, 0.0135, 0.0305, ..., -0.0100, 0.0366, 0.0227]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0054, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0149, -0.0054, 0.0288, ..., 0.0136, 0.0583, 0.0044], + [-0.0281, 0.0154, -0.0386, ..., -0.0444, -0.0093, -0.0009], + ..., + [ 0.0035, 0.0157, 0.0217, ..., 0.0047, 0.0115, 0.0151], + [ 0.0051, 0.0195, -0.0283, ..., -0.0048, -0.0039, -0.0342], + [-0.0167, -0.0195, 0.0007, ..., -0.0048, 0.0165, 0.0132]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0140, -0.0146, 0.0459, ..., 0.0043, -0.0320, 0.0275], + [ 0.0034, 0.0061, 0.0272, ..., -0.0219, -0.0183, -0.0265], + [ 0.0219, 0.0216, -0.0201, ..., -0.0022, 0.0052, 0.0041], + ..., + [-0.0258, 0.0028, 0.0215, ..., 0.0017, 0.0236, -0.0269], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0060], + [-0.0037, -0.0085, -0.0128, ..., -0.0334, 0.0062, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0146, -0.0454, ..., -0.0126, -0.0131, 0.0400], + [ 0.0295, -0.0187, -0.0106, ..., -0.0283, 0.0040, -0.0186], + [ 0.0013, 0.0547, -0.0042, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0280, 0.0045, -0.0269, ..., -0.0090, -0.0287, 0.0150], + [ 0.0093, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0098, 0.0106, ..., -0.0075, 0.0703, -0.0059]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0094, -0.0315, ..., 0.0058, 0.0254, 0.0359], + [ 0.0087, -0.0048, 0.0028, ..., -0.0125, -0.0233, -0.0042], + [ 0.0070, 0.0054, -0.0121, ..., -0.0123, 0.0018, 0.0089], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, 0.0040], + [ 0.0339, -0.0059, 0.0223, ..., 0.0410, 0.0028, 0.0138], + [-0.0055, -0.0439, -0.0486, ..., 0.0140, -0.0130, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0002, 0.0139, -0.0042, ..., -0.0079, -0.0045, -0.0229], + [-0.0289, 0.0231, 0.0084, ..., -0.0140, 0.0056, 0.0222], + [ 0.0085, 0.0295, -0.0193, ..., -0.0164, 0.0310, -0.0262], + ..., + [ 0.0233, -0.0053, -0.0029, ..., 0.0051, 0.0278, 0.0031], + [-0.0104, -0.0018, 0.0140, ..., -0.0175, 0.0444, 0.0425], + [ 0.0054, -0.0084, 0.0096, ..., 0.0061, -0.0344, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[ 0.0027, 0.0126, 0.0188, ..., -0.0092, -0.0199, 0.0147], + [-0.0581, 0.0294, -0.0043, ..., -0.0240, 0.0058, 0.0302], + [ 0.0087, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [ 0.0002, -0.0292, 0.0537, ..., 0.0033, -0.0078, -0.0106], + [-0.0036, -0.0077, 0.0209, ..., 0.0043, 0.0093, 0.0187], + [-0.0087, -0.0349, -0.0173, ..., -0.0400, 0.0057, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0182, -0.0265, -0.0105, ..., 0.0099, -0.0132, -0.0405], + [ 0.0051, -0.0087, 0.0266, ..., -0.0400, 0.0060, 0.0391], + [-0.0016, 0.0359, 0.0284, ..., 0.0292, -0.0233, 0.0212], + ..., + [ 0.0026, 0.0121, 0.0132, ..., -0.0141, -0.0048, 0.0095], + [ 0.0299, -0.0167, -0.0042, ..., -0.0013, -0.0002, -0.0206], + [ 0.0072, -0.0045, -0.0042, ..., 0.0037, -0.0022, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, -0.0015, 0.0118, ..., -0.0029, -0.0208, -0.0074], + [-0.0096, -0.0284, -0.0192, ..., 0.0376, 0.0214, 0.0232], + [ 0.0134, -0.0278, 0.0118, ..., 0.0396, -0.0079, -0.0069], + ..., + [-0.0001, -0.0018, -0.0311, ..., 0.0344, 0.0051, -0.0123], + [-0.0236, -0.0102, -0.0072, ..., 0.0332, -0.0114, -0.0179], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0137, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[-0.0014, 0.0166, 0.0036, ..., 0.0273, -0.0352, 0.0156], + [-0.0510, -0.0062, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0077, ..., -0.0388, -0.0019, 0.0398], + ..., + [-0.0160, -0.0195, -0.0027, ..., 0.0037, -0.0056, 0.0029], + [-0.0187, 0.0038, 0.0098, ..., 0.0287, -0.0125, -0.0095], + [-0.0181, -0.0032, 0.0247, ..., -0.0143, 0.0052, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0060, 0.0096, 0.0141, ..., -0.0112, 0.0026, -0.0253], + [-0.0167, 0.0330, -0.0171, ..., -0.0037, -0.0403, -0.0110], + [ 0.0084, 0.0347, 0.0275, ..., 0.0148, 0.0064, 0.0079], + ..., + [ 0.0272, -0.0168, -0.0106, ..., -0.0098, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0059, ..., -0.0032, -0.0018, -0.0243], + [-0.0075, 0.0581, 0.0057, ..., 0.0145, 0.0239, 0.0116]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0118, -0.0410, 0.0126, ..., -0.0254, -0.0052, -0.0173], + [ 0.0476, 0.0315, 0.0148, ..., -0.0150, -0.0066, -0.0012], + [-0.0089, -0.0269, 0.0083, ..., 0.0332, 0.0144, -0.0063], + ..., + [-0.0046, -0.0043, 0.0160, ..., -0.0177, 0.0168, -0.0126], + [ 0.0051, -0.0187, -0.0049, ..., 0.0248, -0.0148, -0.0068], + [-0.0142, -0.0024, 0.0148, ..., -0.0043, 0.0037, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0054, 0.0094, ..., 0.0255, -0.0253, -0.0135], + [-0.0018, -0.0143, -0.0160, ..., -0.0113, -0.0403, -0.0408], + [ 0.0176, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0115, 0.0046, 0.0322, ..., -0.0183, -0.0010, 0.0016], + [ 0.0034, -0.0050, 0.0162, ..., 0.0043, 0.0255, 0.0063], + [ 0.0270, 0.0087, 0.0159, ..., 0.0189, 0.0148, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0066, ..., -0.0532, 0.0104, 0.0036], + [-0.0240, -0.0215, -0.0103, ..., -0.0059, 0.0098, -0.0168], + [ 0.0074, -0.0013, -0.0059, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0057, 0.0172, -0.0106, ..., -0.0300, -0.0045, 0.0435], + [ 0.0051, -0.0261, 0.0247, ..., 0.0192, -0.0308, 0.0061], + [-0.0167, 0.0123, 0.0106, ..., -0.0247, 0.0123, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0078, -0.0078, 0.0317, ..., 0.0200, -0.0104, -0.0376], + [ 0.0356, 0.0378, 0.0135, ..., 0.0325, 0.0391, 0.0173], + [-0.0244, -0.0264, 0.0371, ..., 0.0028, 0.0052, -0.0322], + ..., + [ 0.0058, 0.0107, -0.0040, ..., 0.0366, -0.0206, 0.0374], + [-0.0046, -0.0070, 0.0034, ..., -0.0093, 0.0090, 0.0192], + [ 0.0104, -0.0178, -0.0026, ..., -0.0166, -0.0056, 0.0245]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0053, -0.0093, -0.0344, ..., -0.0076, 0.0347, -0.0176], + [ 0.0267, -0.0366, -0.0088, ..., 0.0217, 0.0349, -0.0043], + [-0.0097, -0.0286, 0.0074, ..., 0.0052, -0.0282, 0.0013], + ..., + [-0.0129, 0.0037, 0.0058, ..., -0.0086, 0.0165, 0.0076], + [-0.0322, -0.0085, -0.0170, ..., 0.0195, 0.0212, -0.0129], + [-0.0140, 0.0043, -0.0170, ..., -0.0127, 0.0060, 0.0070]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0025, -0.0047, ..., -0.0483, 0.0374, 0.0093], + [-0.0576, 0.0116, 0.0374, ..., -0.0153, -0.0188, -0.0136], + ..., + [ 0.0260, -0.0007, -0.0077, ..., -0.0003, 0.0053, -0.0165], + [-0.0222, -0.0155, -0.0437, ..., -0.0066, -0.0165, -0.0282], + [-0.0131, -0.0148, -0.0038, ..., 0.0018, -0.0234, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0146, -0.0649, 0.0101], + [ 0.0249, 0.0205, 0.0075, ..., -0.0061, -0.0089, -0.0105], + [-0.0220, 0.0044, 0.0047, ..., 0.0309, 0.0381, 0.0037], + ..., + [ 0.0200, -0.0033, 0.0171, ..., 0.0122, 0.0320, -0.0157], + [ 0.0349, 0.0120, 0.0195, ..., 0.0398, 0.0315, -0.0217], + [-0.0075, 0.0256, 0.0050, ..., 0.0205, 0.0035, -0.0105]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0043, 0.0199, ..., -0.0050, 0.0276, -0.0308], + [ 0.0178, -0.0063, 0.0099, ..., 0.0108, -0.0398, 0.0300], + [ 0.0430, -0.0124, 0.0081, ..., -0.0009, -0.0066, -0.0055], + ..., + [ 0.0192, -0.0106, -0.0108, ..., -0.0025, 0.0330, 0.0142], + [-0.0124, -0.0542, -0.0205, ..., 0.0062, -0.0159, 0.0320], + [ 0.0303, 0.0054, -0.0425, ..., 0.0009, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0068, -0.0315, ..., 0.0525, 0.0034, -0.0349], + [-0.0044, -0.0070, -0.0156, ..., 0.0053, -0.0240, 0.0305], + [-0.0256, -0.0045, 0.0062, ..., -0.0459, -0.0148, -0.0165], + ..., + [ 0.0070, 0.0231, 0.0209, ..., 0.0077, -0.0038, 0.0461], + [-0.0067, -0.0085, 0.0004, ..., -0.0217, 0.0347, -0.0018], + [ 0.0057, 0.0234, 0.0056, ..., 0.0261, -0.0038, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0071, 0.0023], + [-0.0183, -0.0211, 0.0045, ..., -0.0145, 0.0143, -0.0166], + [-0.0075, -0.0160, 0.0053, ..., -0.0203, 0.0267, 0.0217], + ..., + [ 0.0308, 0.0096, 0.0075, ..., 0.0171, 0.0053, -0.0070], + [-0.0145, 0.0260, -0.0085, ..., 0.0077, 0.0102, -0.0166], + [ 0.0094, -0.0142, -0.0037, ..., 0.0051, 0.0069, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0070, 0.0130, -0.0076, ..., -0.0278, 0.0240, -0.0126], + [ 0.0035, 0.0187, -0.0160, ..., -0.0317, -0.0032, 0.0173], + [ 0.0062, 0.0310, -0.0092, ..., -0.0226, 0.0030, 0.0334], + ..., + [-0.0079, 0.0186, 0.0038, ..., 0.0228, 0.0069, 0.0065], + [ 0.0271, 0.0117, 0.0354, ..., -0.0076, 0.0198, 0.0145], + [ 0.0282, 0.0171, 0.0317, ..., -0.0111, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0114, -0.0265, -0.0110, ..., 0.0049, 0.0037, 0.0136], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0085], + [-0.0067, -0.0190, -0.0045, ..., 0.0166, -0.0135, 0.0138], + ..., + [-0.0200, -0.0012, 0.0061, ..., -0.0339, 0.0361, -0.0025], + [ 0.0037, -0.0250, 0.0153, ..., 0.0006, -0.0162, 0.0220], + [-0.0078, 0.0028, 0.0205, ..., -0.0064, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0053, 0.0240, -0.0217, ..., 0.0172, 0.0048, 0.0114], + [-0.0386, 0.0103, 0.0075, ..., -0.0042, 0.0322, 0.0086], + [-0.0427, 0.0040, 0.0035, ..., 0.0073, 0.0110, -0.0137], + ..., + [-0.0087, -0.0041, -0.0264, ..., -0.0586, -0.0630, -0.0132], + [ 0.0045, 0.0287, 0.0420, ..., -0.0311, -0.0090, -0.0289], + [-0.0107, -0.0610, 0.0162, ..., -0.0291, -0.0166, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0121, -0.0251, 0.0211, ..., 0.0199, -0.0084, -0.0112], + [ 0.0112, -0.0099, -0.0327, ..., 0.0447, 0.0076, 0.0074], + [ 0.0035, -0.0175, 0.0066, ..., -0.0062, -0.0049, -0.0065], + ..., + [ 0.0295, 0.0045, -0.0104, ..., 0.0093, -0.0160, -0.0288], + [ 0.0083, 0.0253, -0.0177, ..., 0.0094, 0.0074, 0.0080], + [-0.0129, 0.0027, 0.0137, ..., -0.0128, 0.0277, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0059, -0.0276, -0.0271], + [ 0.0131, 0.0047, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0068, 0.0234, 0.0065, ..., 0.0464, 0.0064, 0.0309], + ..., + [-0.0117, -0.0175, 0.0327, ..., 0.0132, 0.0044, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, 0.0059, -0.0444, ..., 0.0060, -0.0283, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0082, -0.0099, 0.0146, ..., -0.0742, -0.0064, 0.0161], + [-0.0076, -0.0131, -0.0075, ..., 0.0064, 0.0026, 0.0238], + [ 0.0189, 0.0221, -0.0088, ..., 0.0050, 0.0408, 0.0052], + ..., + [ 0.0059, 0.0247, 0.0396, ..., 0.0064, -0.0304, -0.0183], + [ 0.0299, -0.0114, -0.0417, ..., -0.0048, -0.0376, -0.0044], + [-0.0210, 0.0168, -0.0026, ..., 0.0049, 0.0178, 0.0092]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[ 0.0069, 0.0242, 0.0062, ..., -0.0245, 0.0078, 0.0002], + [ 0.0190, -0.0125, 0.0332, ..., -0.0146, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0068, ..., -0.0059, -0.0092, 0.0182], + ..., + [-0.0430, -0.0403, -0.0081, ..., 0.0135, -0.0140, 0.0046], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0094, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0042, 0.0121, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0120, ..., -0.0160, -0.0272, 0.0299], + [ 0.0092, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0067, ..., 0.0101, -0.0223, 0.0062], + ..., + [ 0.0192, 0.0292, 0.0059, ..., 0.0176, -0.0197, 0.0046], + [ 0.0265, 0.0063, 0.0054, ..., 0.0286, -0.0021, -0.0267], + [-0.0184, -0.0088, -0.0048, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0084, -0.0159, 0.0082, ..., 0.0198, 0.0056, 0.0277], + [ 0.0172, 0.0041, -0.0057, ..., 0.0048, -0.0049, 0.0057], + [ 0.0036, 0.0184, 0.0113, ..., 0.0099, -0.0056, -0.0074], + ..., + [ 0.0048, -0.0111, 0.0035, ..., -0.0096, -0.0053, 0.0122], + [-0.0427, -0.0391, -0.0192, ..., 0.0091, -0.0238, -0.0078], + [ 0.0055, -0.0099, -0.0222, ..., -0.0164, 0.0342, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0084, -0.0161, 0.0123], + [-0.0081, 0.0060, 0.0312, ..., 0.0053, 0.0286, 0.0317], + [ 0.0215, -0.0142, 0.0232, ..., -0.0071, -0.0014, 0.0278], + ..., + [ 0.0170, -0.0085, 0.0471, ..., 0.0067, 0.0078, -0.0073], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0076, -0.0203], + [ 0.0053, 0.0048, -0.0081, ..., -0.0095, -0.0086, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[-0.0058, -0.0156, 0.0284, ..., -0.0104, 0.0135, -0.0044], + [ 0.0152, 0.0332, 0.0337, ..., 0.0223, -0.0165, -0.0044], + [ 0.0055, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [-0.0056, -0.0019, 0.0280, ..., -0.0076, -0.0078, 0.0140], + [-0.0083, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0091, ..., 0.0067, 0.0167, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0042, ..., -0.0238, -0.0103, 0.0119], + [-0.0250, -0.0162, 0.0034, ..., -0.0081, 0.0002, -0.0369], + [ 0.0051, 0.0141, 0.0079, ..., -0.0452, -0.0038, 0.0120], + ..., + [ 0.0148, -0.0141, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0101, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0142, 0.0041, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[ 0.0028, -0.0327, -0.0190, ..., 0.0064, -0.0052, -0.0209], + [ 0.0713, 0.0152, -0.0052, ..., -0.0049, 0.0186, 0.0027], + [-0.0227, -0.0092, 0.0188, ..., -0.0122, 0.0193, 0.0053], + ..., + [ 0.0116, 0.0464, 0.0085, ..., 0.0248, 0.0168, 0.0047], + [ 0.0381, -0.0038, 0.0071, ..., -0.0031, -0.0107, -0.0054], + [ 0.0292, 0.0057, 0.0209, ..., 0.0056, 0.0234, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0056, 0.0212, ..., 0.0305, 0.0137, 0.0105], + [-0.0078, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0045], + ..., + [ 0.0084, -0.0275, 0.0225, ..., 0.0043, -0.0189, -0.0107], + [-0.0128, 0.0053, -0.0058, ..., -0.0079, 0.0064, 0.0211], + [ 0.0015, -0.0073, -0.0063, ..., -0.0109, 0.0264, 0.0059]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[-0.0049, -0.0075, -0.0108, ..., 0.0186, -0.0216, -0.0047], + [ 0.0048, 0.0123, -0.0231, ..., 0.0123, -0.0112, -0.0183], + [ 0.0107, 0.0103, -0.0057, ..., 0.0425, 0.0309, -0.0273], + ..., + [ 0.0332, 0.0125, 0.0154, ..., 0.0540, 0.0415, 0.0097], + [ 0.0025, 0.0087, -0.0137, ..., -0.0030, -0.0189, -0.0237], + [ 0.0059, 0.0354, 0.0177, ..., 0.0361, -0.0311, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0076, 0.0359, ..., -0.0084, -0.0120, -0.0049], + [ 0.0190, -0.0059, 0.0298, ..., 0.0126, -0.0050, -0.0200], + [ 0.0112, -0.0110, 0.0177, ..., -0.0248, 0.0229, 0.0034], + ..., + [-0.0081, 0.0044, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, 0.0044, -0.0291, ..., 0.0206, 0.0304, -0.0085], + [-0.0080, 0.0245, -0.0068, ..., -0.0413, 0.0216, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0078, -0.0157, 0.0043, ..., -0.0052, 0.0228, -0.0339], + [-0.0713, -0.0153, -0.0168, ..., -0.0217, -0.0183, -0.0073], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, -0.0063], + ..., + [-0.0027, -0.0069, 0.0258, ..., -0.0310, 0.0091, -0.0361], + [ 0.0063, 0.0308, 0.0542, ..., -0.0542, -0.0156, 0.0259], + [-0.0071, 0.0157, -0.0069, ..., 0.0075, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0077, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0106, ..., 0.0104, -0.0210, 0.0148], + [-0.0286, -0.0267, -0.0037, ..., 0.0129, -0.0059, 0.0087], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [-0.0062, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0129, 0.0206, ..., 0.0069, -0.0121, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0152, -0.0092, -0.0231, ..., -0.0214, 0.0447, 0.0135], + [-0.0171, -0.0045, 0.0076, ..., -0.0046, 0.0110, 0.0598], + [-0.0062, 0.0267, -0.0053, ..., 0.0284, -0.0045, -0.0364], + ..., + [ 0.0045, 0.0078, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0017, -0.0087, ..., 0.0211, 0.0137, -0.0075], + [ 0.0101, 0.0236, -0.0206, ..., -0.0136, -0.0063, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0067, 0.0294, ..., 0.0040, 0.0270, 0.0065], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0066, -0.0168], + [ 0.0201, -0.0121, 0.0126, ..., -0.0381, -0.0247, -0.0059], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0078], + [-0.0194, 0.0094, -0.0092, ..., -0.0231, -0.0069, -0.0466], + [ 0.0327, 0.0284, -0.0044, ..., 0.0145, 0.0142, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0137, -0.0021, -0.0167, ..., -0.0037, 0.0170, 0.0107], + [ 0.0019, -0.0031, 0.0295, ..., -0.0172, 0.0040, 0.0135], + [ 0.0193, 0.0236, -0.0125, ..., 0.0038, -0.0369, 0.0239], + ..., + [ 0.0151, -0.0413, -0.0124, ..., -0.0354, 0.0065, -0.0320], + [-0.0266, -0.0063, -0.0327, ..., -0.0050, 0.0058, 0.0073], + [ 0.0298, 0.0356, -0.0192, ..., 0.0042, 0.0008, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0045, 0.0058, 0.0317, ..., 0.0062, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0071, -0.0096], + [-0.0166, 0.0064, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0041, 0.0260, ..., 0.0159, -0.0035, -0.0474], + [-0.0064, 0.0079, 0.0076, ..., 0.0140, 0.0557, 0.0052], + [-0.0203, 0.0273, -0.0040, ..., -0.0330, -0.0038, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-3.3203e-02, -1.2390e-02, -3.8757e-03, ..., 6.3782e-03, + -9.3384e-03, -2.7710e-02], + [ 4.9744e-03, -3.4637e-03, -5.5237e-03, ..., 6.5002e-03, + 1.4038e-02, -1.6113e-02], + [-4.6997e-03, 1.6846e-02, -7.6599e-03, ..., -9.1553e-03, + 1.2390e-02, 2.0447e-03], + ..., + [ 3.9795e-02, 4.5166e-02, 2.5146e-02, ..., -5.8289e-03, + -3.2959e-02, 4.1199e-03], + [-6.2012e-02, 3.5248e-03, 3.7109e-02, ..., 4.1016e-02, + 9.2506e-05, -1.7334e-02], + [ 1.2634e-02, 5.7678e-03, -2.6978e-02, ..., 4.9438e-03, + 3.8818e-02, -1.0376e-02]], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0126, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0013, 0.0245], + [-0.0449, 0.0019, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, 0.0062, ..., 0.0041, 0.0288, 0.0124], + [ 0.0330, 0.0038, 0.0060, ..., 0.0299, 0.0142, -0.0483], + [-0.0166, 0.0077, 0.0051, ..., -0.0062, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0045, -0.0025, 0.0527], + [-0.0002, 0.0403, 0.0112, ..., 0.0417, 0.0105, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0096, -0.0096, 0.0327], + [ 0.0142, 0.0111, -0.0069, ..., 0.0090, -0.0178, -0.0254], + [-0.0530, -0.0124, 0.0281, ..., -0.0242, -0.0127, -0.0047]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, -0.0042], + [ 0.0063, -0.0045, -0.0286, ..., 0.0056, 0.0172, 0.0106], + [ 0.0393, 0.0400, 0.0039, ..., 0.0025, 0.0082, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0107, 0.0097, 0.0150], + [-0.0057, 0.0039, 0.0077, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0084, 0.0078, ..., 0.0092, -0.0114, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0071, 0.0041, ..., -0.0220, -0.0145, -0.0013], + [-0.0166, 0.0189, -0.0047, ..., 0.0096, 0.0020, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0054, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0057, ..., 0.0047, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0082, 0.0071, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0122, 0.0289, 0.0125, ..., 0.0135, 0.0070, -0.0104], + [ 0.0188, -0.0197, 0.0010, ..., 0.0284, -0.0342, -0.0057], + [-0.0525, -0.0215, 0.0154, ..., -0.0270, 0.0081, 0.0254], + ..., + [-0.0069, 0.0087, 0.0281, ..., 0.0129, -0.0064, -0.0048], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0045, 0.0038], + [-0.0273, -0.0747, 0.0208, ..., -0.0044, -0.0044, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0048, 0.0150, 0.0334, ..., 0.0205, -0.0273, -0.0037], + [-0.0165, 0.0111, -0.0135, ..., 0.0106, -0.0071, 0.0219], + [ 0.0063, 0.0227, -0.0054, ..., 0.0128, -0.0077, 0.0150], + ..., + [-0.0347, -0.0148, 0.0053, ..., -0.0046, -0.0042, -0.0094], + [-0.0077, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0069, -0.0500, -0.0232, ..., 0.0320, 0.0057, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0056, -0.0364], + [ 0.0220, -0.0026, -0.0112, ..., -0.0052, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0115, ..., 0.0444, 0.0039, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0064, 0.0192, -0.0393], + [-0.0102, -0.0244, 0.0275, ..., 0.0029, 0.0042, 0.0081], + [-0.0170, -0.0427, 0.0140, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0040, -0.0378, -0.0359, ..., 0.0552, -0.0062, -0.0160], + [ 0.0039, 0.0143, 0.0134, ..., 0.0049, -0.0153, 0.0354], + [ 0.0095, -0.0038, 0.0065, ..., -0.0046, -0.0249, -0.0136], + ..., + [ 0.0081, -0.0124, 0.0050, ..., -0.0095, -0.0085, 0.0047], + [-0.0051, 0.0439, 0.0166, ..., 0.0248, 0.0131, -0.0187], + [ 0.0175, 0.0059, 0.0041, ..., 0.0410, -0.0096, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0104, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0141, -0.0050, ..., 0.0052, -0.0244, 0.0356], + [ 0.0063, 0.0165, -0.0200, ..., -0.0019, 0.0237, -0.0078], + ..., + [-0.0197, -0.0049, 0.0386, ..., 0.0181, 0.0112, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, 0.0041, -0.0374], + [-0.0086, -0.0096, 0.0172, ..., 0.0243, -0.0220, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-0.0347, 0.0066, 0.0193, ..., -0.0221, -0.0150, -0.0211], + [ 0.0135, 0.0114, 0.0253, ..., 0.0204, -0.0232, -0.0322], + [ 0.0255, 0.0045, -0.0041, ..., 0.0050, -0.0055, 0.0337], + ..., + [ 0.0422, 0.0033, 0.0352, ..., -0.0304, -0.0049, -0.0204], + [ 0.0302, 0.0156, 0.0099, ..., -0.0078, -0.0254, -0.0245], + [-0.0040, -0.0060, -0.0039, ..., 0.0037, -0.0197, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0052, 0.0266, ..., 0.0154, 0.0187, 0.0047], + [-0.0119, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0051, ..., -0.0172, -0.0091, 0.0231], + ..., + [-0.0295, -0.0137, 0.0038, ..., 0.0043, -0.0044, 0.0099], + [-0.0049, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0078], + [-0.0040, -0.0048, -0.0259, ..., -0.0039, -0.0140, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0079, 0.0078, -0.0140, ..., 0.0242, -0.0070, -0.0386], + [-0.0140, 0.0079, -0.0097, ..., -0.0369, -0.0156, 0.0157], + [-0.0267, 0.0078, -0.0057, ..., -0.0126, 0.0183, -0.0120], + ..., + [ 0.0172, 0.0007, 0.0144, ..., 0.0115, 0.0042, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., -0.0046, 0.0042, -0.0137], + [-0.0219, -0.0396, 0.0041, ..., -0.0058, -0.0018, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-0.0034, 0.0168, 0.0194, ..., 0.0349, -0.0420, 0.0017], + [ 0.0042, 0.0273, 0.0145, ..., -0.0184, -0.0039, 0.0154], + [ 0.0145, -0.0036, -0.0072, ..., 0.0245, 0.0359, 0.0059], + ..., + [ 0.0432, 0.0145, -0.0320, ..., 0.0046, 0.0033, 0.0266], + [ 0.0229, 0.0126, -0.0199, ..., -0.0070, -0.0312, -0.0457], + [ 0.0050, 0.0054, -0.0053, ..., 0.0049, -0.0107, -0.0156]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0114, 0.0388, 0.0359], + [-0.0015, 0.0123, -0.0029, ..., 0.0124, -0.0038, 0.0066], + [-0.0277, -0.0036, -0.0181, ..., -0.0027, -0.0034, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0042, 0.0513, -0.0110], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0111], + [-0.0282, -0.0053, 0.0121, ..., -0.0051, -0.0039, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0039, -0.0128, ..., 0.0140, 0.0308, 0.0058], + [ 0.0119, -0.0332, -0.0320, ..., -0.0078, -0.0140, -0.0029], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, -0.0028, -0.0311, ..., 0.0120, 0.0061, 0.0081], + [ 0.0048, 0.0198, 0.0142, ..., -0.0222, -0.0144, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, -0.0046, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0091, -0.0036, 0.0056, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0018, ..., 0.0302, -0.0024, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0039, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0025, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0049, -0.0222, 0.0262], + ..., + [-0.0147, 0.0058, 0.0260, ..., 0.0051, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0057, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0076, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0113], + [-0.0278, 0.0044, -0.0374, ..., 0.0276, -0.0038, 0.0151], + [ 0.0089, 0.0237, -0.0067, ..., 0.0527, -0.0070, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0068, -0.0095], + [-0.0075, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0044, -0.0128, ..., 0.0049, -0.0371, 0.0017], + [ 0.0075, -0.0126, 0.0049, ..., -0.0204, 0.0041, 0.0056], + [-0.0110, -0.0195, 0.0056, ..., -0.0542, -0.0092, -0.0053], + ..., + [-0.0205, -0.0085, -0.0041, ..., -0.0199, 0.0041, 0.0061], + [-0.0220, -0.0302, -0.0099, ..., 0.0104, 0.0369, 0.0043], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0135, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[-0.0023, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0047, 0.0049, 0.0238, ..., -0.0145, -0.0320, 0.0037], + [-0.0193, 0.0364, -0.0048, ..., -0.0064, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0048, -0.0232, -0.0464], + [-0.0109, -0.0298, -0.0195, ..., 0.0114, 0.0312, 0.0156], + [-0.0074, -0.0117, -0.0195, ..., 0.0043, 0.0106, 0.0099]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0151, ..., -0.0371, 0.0400, -0.0165], + [ 0.0048, -0.0239, -0.0123, ..., -0.0045, 0.0243, -0.0396], + [-0.0101, 0.0155, -0.0056, ..., -0.0277, -0.0038, 0.0037], + ..., + [ 0.0117, -0.0057, 0.0515, ..., 0.0327, -0.0120, -0.0066], + [ 0.0041, 0.0170, 0.0074, ..., -0.0048, 0.0280, -0.0210], + [ 0.0036, 0.0068, 0.0057, ..., 0.0153, -0.0077, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0061, 0.0081, -0.0077, ..., -0.0588, 0.0187, -0.0064], + [-0.0102, 0.0199, -0.0077, ..., 0.0080, -0.0332, 0.0040], + ..., + [ 0.0078, -0.0464, 0.0277, ..., -0.0145, 0.0198, -0.0078], + [ 0.0162, 0.0038, 0.0090, ..., -0.0109, 0.0081, 0.0297], + [-0.0269, 0.0135, 0.0164, ..., 0.0079, 0.0176, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0042, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, -0.0026, ..., 0.0142, 0.0127, 0.0439], + [-0.0352, 0.0048, -0.0201, ..., 0.0330, 0.0309, -0.0051], + ..., + [ 0.0039, 0.0123, 0.0153, ..., 0.0039, -0.0039, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0040, 0.0175], + [-0.0244, -0.0059, -0.0300, ..., 0.0049, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0045, -0.0041, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0038, ..., 0.0310, 0.0088, -0.0039], + [ 0.0025, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0110, -0.0237, ..., -0.0104, 0.0157, 0.0078], + [-0.0070, 0.0205, 0.0143, ..., 0.0262, -0.0195, 0.0095], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0078]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, -0.0024, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0022, 0.0034, -0.0093], + [ 0.0017, -0.0010, 0.0398, ..., -0.0304, -0.0028, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0020], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0063, 0.0228, ..., -0.0322, 0.0315, -0.0039], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0051, 0.0291], + [ 0.0132, 0.0038, -0.0339, ..., 0.0077, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, -0.0027], + [-0.0166, -0.0036, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0061, -0.0032, ..., 0.0240, 0.0036, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, 0.0019], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0017, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0007, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0067, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0040], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, 0.0040, -0.0039], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, -0.0039, ..., -0.0164, -0.0039, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0039, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0026, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0025, 0.0226], + ..., + [ 0.0053, -0.0231, -0.0255, ..., -0.0052, 0.0295, -0.0058], + [-0.0255, 0.0513, 0.0039, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0040, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0056, 0.0078, -0.0364, ..., -0.0175, 0.0306, -0.0075], + [-0.0051, -0.0195, -0.0016, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0070, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0078, -0.0193, -0.0140], + [-0.0166, 0.0059, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0138, 0.0082, 0.0146, ..., -0.0081, 0.0211, -0.0270], + [ 0.0063, -0.0077, -0.0234, ..., 0.0089, -0.0055, -0.0165], + [ 0.0225, -0.0251, 0.0047, ..., 0.0072, 0.0236, -0.0036], + ..., + [ 0.0186, 0.0148, -0.0053, ..., 0.0272, -0.0062, -0.0086], + [ 0.0064, 0.0029, 0.0078, ..., -0.0260, 0.0267, 0.0147], + [ 0.0101, -0.0044, 0.0212, ..., 0.0195, -0.0157, 0.0114]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0084, -0.0270, 0.0251, ..., 0.0098, 0.0055, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.2.proj.1.weight': tensor([[ 0.0036, 0.0231, 0.0918, ..., -0.0571, 0.0250, -0.0284], + [-0.0237, 0.0006, -0.1128, ..., 0.0193, -0.0238, 0.0518], + [-0.0087, 0.1279, -0.0123, ..., -0.0220, 0.0184, -0.0349], + ..., + [ 0.0146, 0.0055, -0.0076, ..., 0.0234, -0.0280, 0.0082], + [-0.0237, -0.0608, 0.0603, ..., 0.0332, 0.0708, 0.0505], + [ 0.0146, -0.0229, -0.0050, ..., 0.0249, 0.0420, -0.0055]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_proj.weight': tensor([[-0.0093, 0.0068, 0.0118, ..., -0.0078, -0.0047, 0.0182], + [ 0.0018, 0.0015, -0.0032, ..., -0.0032, -0.0005, -0.0092], + [-0.0064, 0.0067, -0.0002, ..., 0.0002, -0.0104, -0.0227], + ..., + [ 0.0010, -0.0172, -0.0045, ..., 0.0082, 0.0030, -0.0064], + [-0.0014, 0.0045, -0.0045, ..., -0.0121, -0.0164, -0.0048], + [ 0.0080, -0.0050, -0.0016, ..., 0.0063, 0.0061, 0.0010]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_norm.weight': tensor([ 1.4062e-01, -4.2480e-02, 8.6426e-02, 2.3145e-01, -3.4485e-03, + -8.3160e-04, -1.3933e-06, -1.8794e-21, 6.6174e-22, -2.0117e-20, + -9.3408e-33, 3.3352e-21, -6.1512e-05, -1.1942e-23, -2.9565e-25, + 2.0954e-30, 1.8597e-05, 2.6484e-09, -3.7517e-12, -1.8406e-04, + 2.4738e-10, -4.5528e-21, 4.1504e-02, -3.2616e-04, -1.0742e-01, + 5.8594e-01, 5.4932e-02, -7.3853e-03, 4.9805e-02, 9.8633e-02, + -6.3019e-19, -1.4095e-21, 1.0757e-03, -9.5367e-05, 1.2346e-13, + -2.4033e-04, 3.4106e-12, -7.1086e-25, 1.1529e-23, -2.0292e-24, + -2.3588e-25, -6.0988e-26, 1.1228e-25, -6.5308e-03, 4.2969e-02, + -1.4771e-02, 2.9785e-02, 1.2354e-01, 3.4180e-01, 7.8125e-03, + -4.6082e-03, 5.3406e-03, 4.4556e-03, 1.2005e-10, 2.8922e-10, + 1.4095e-17, -4.8307e-22, -1.0503e-18, 3.0494e-26, -2.6496e-24, + 5.1670e-29, -8.8938e-20, 2.1487e-25, 2.4536e-14, 2.7148e-01, + 2.2559e-01, 3.7109e-01, -4.9316e-02, 2.6855e-03, 7.1335e-04, + 1.0282e-06, 4.3675e-22, 1.2811e-20, 5.7175e-20, -1.3482e-31, + -7.3123e-22, 3.6163e-03, -7.3412e-24, -2.1306e-26, 6.2320e-29, + 1.9226e-03, -9.7316e-11, 4.0927e-11, -3.8452e-03, 3.3253e-12, + -6.6746e-19, -3.9978e-03, -1.2360e-03, 5.7812e-01, 1.1719e-01, + 5.0391e-01, 9.2773e-02, -5.8594e-03, 9.0408e-04, 1.6263e-18, + -5.2675e-21, 3.0060e-03, 4.6730e-05, 4.2188e-14, 3.9978e-03, + -5.7116e-10, -1.4217e-24, -8.4011e-25, -2.8290e-22, -4.2914e-28, + -2.5022e-23, 8.5591e-29, 1.4404e-02, 5.4626e-03, -6.0425e-03, + 4.1211e-01, 2.7734e-01, 2.0410e-01, -1.1047e-02, -5.2185e-03, + 7.8125e-03, 4.6692e-03, -4.8885e-11, 1.4461e-10, -9.1480e-19, + 4.1558e-21, 1.2197e-18, 3.3120e-26, -3.9911e-23, -1.2572e-29, + -3.2591e-22, -1.7707e-24, 8.5038e-11], dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.self_attn.k_proj.weight': tensor([[-0.0053, 0.0036, -0.0114, ..., -0.0015, -0.0072, 0.0092], + [-0.0233, -0.0015, 0.0042, ..., -0.0058, -0.0069, -0.0112], + [-0.0036, -0.0111, -0.0016, ..., 0.0052, 0.0021, 0.0089], + ..., + [ 0.0049, -0.0039, -0.0018, ..., -0.0025, 0.0008, 0.0016], + [ 0.0049, 0.0009, 0.0007, ..., -0.0044, 0.0020, 0.0024], + [ 0.0040, 0.0023, 0.0050, ..., -0.0034, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.k_norm.weight': tensor([ 2.4121e-01, 1.2598e-01, 6.2012e-02, 2.0898e-01, 3.3417e-03, + -1.1292e-03, 1.4082e-06, -1.9852e-21, 2.2896e-21, 2.8164e-20, + 1.9548e-32, 3.2955e-21, -9.9659e-05, -1.1581e-23, -2.9404e-25, + 2.7364e-30, -7.3016e-07, 2.6921e-09, 2.0464e-12, 1.9360e-04, + -2.2555e-10, 2.7926e-21, 3.7354e-02, -9.9182e-04, 7.6172e-02, + 5.8594e-01, 2.2852e-01, 8.9844e-02, 6.2256e-03, 6.8970e-03, + 2.5072e-19, 9.3306e-22, -3.7003e-04, -9.8228e-05, 1.4122e-13, + -1.0824e-04, -3.0695e-11, -1.3248e-25, -1.1684e-23, 1.0792e-24, + -2.2941e-25, -1.1551e-25, -1.1228e-25, 1.0681e-02, 4.4678e-02, + -1.1536e-02, 2.2852e-01, 2.0996e-01, 3.7109e-01, 1.2085e-02, + -4.2725e-03, -7.8125e-03, 7.6904e-03, -1.2824e-10, 3.1650e-10, + -7.5352e-18, -1.7999e-21, 1.3281e-18, 4.2005e-26, -2.7271e-24, + -5.4037e-29, -8.6821e-20, -2.1487e-25, 2.6867e-14, -1.0547e-01, + 2.2168e-01, 3.7305e-01, -1.5430e-01, -2.1667e-03, -9.6321e-05, + -1.0431e-06, -6.3693e-23, 1.2705e-20, 5.5904e-20, 1.3173e-31, + 1.0059e-21, -3.5858e-03, 6.2814e-24, -2.0700e-26, 6.6659e-29, + 1.9302e-03, 2.0191e-10, 4.2292e-11, -3.8452e-03, -2.7427e-12, + -6.5391e-19, -7.8735e-03, -2.8229e-04, 5.7812e-01, -1.3477e-01, + 4.9609e-01, 7.0801e-03, 4.9072e-02, 9.4727e-02, -1.8431e-18, + -5.2675e-21, -3.1891e-03, 2.1338e-05, -1.9207e-14, -3.9978e-03, + 5.8208e-10, -1.4282e-24, 3.2829e-24, -2.7132e-22, 1.6787e-27, + -1.5303e-23, 2.9819e-28, 9.4604e-03, 2.0294e-03, 1.0559e-02, + 4.0430e-01, 2.3340e-01, 1.8555e-01, 8.2397e-03, 5.9814e-03, + 4.6387e-03, -1.2054e-03, 1.6257e-11, 2.6489e-11, -8.4026e-18, + 3.0308e-21, -2.2192e-19, -2.1104e-26, 3.8877e-23, -1.0748e-29, + 6.0880e-21, -1.8482e-24, -8.4583e-11], dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.self_attn.v_proj.weight': tensor([[-0.0007, 0.0037, -0.0014, ..., 0.0045, 0.0013, -0.0028], + [ 0.0083, 0.0014, -0.0035, ..., -0.0039, 0.0026, 0.0009], + [ 0.0136, 0.0058, 0.0026, ..., -0.0038, -0.0015, 0.0028], + ..., + [-0.0036, -0.0049, 0.0043, ..., 0.0056, -0.0034, -0.0077], + [-0.0043, -0.0018, -0.0001, ..., 0.0055, -0.0032, -0.0046], + [-0.0015, 0.0055, -0.0029, ..., 0.0037, 0.0025, -0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.self_attn.output_proj.weight': tensor([[ 8.4229e-03, -3.6133e-02, -1.4099e-02, ..., -6.5918e-03, + 5.1270e-03, 5.0659e-03], + [ 6.6223e-03, -7.2327e-03, 5.8594e-03, ..., -2.6703e-03, + 1.8311e-03, 3.2043e-03], + [-3.2196e-03, 5.2214e-05, 7.5684e-03, ..., 4.5776e-03, + 4.5166e-03, 2.8839e-03], + ..., + [ 2.7008e-03, 5.2490e-03, 7.0801e-03, ..., -4.4250e-03, + -6.4697e-03, -6.5002e-03], + [-2.2125e-03, 5.1270e-03, -6.3477e-03, ..., -9.4604e-03, + 8.1787e-03, 4.3640e-03], + [-4.2114e-03, 1.0132e-02, 3.9368e-03, ..., 4.6997e-03, + 4.4861e-03, 4.4556e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_proj.weight': tensor([[ 0.0085, 0.0082, -0.0034, ..., -0.0069, 0.0062, -0.0014], + [ 0.0073, -0.0261, -0.0047, ..., 0.0066, -0.0082, 0.0026], + [-0.0121, 0.0070, -0.0002, ..., 0.0057, 0.0209, -0.0048], + ..., + [ 0.0033, 0.0048, -0.0010, ..., -0.0012, -0.0046, 0.0108], + [-0.0066, -0.0033, 0.0006, ..., 0.0024, 0.0063, -0.0033], + [-0.0024, -0.0065, 0.0020, ..., -0.0033, -0.0007, 0.0029]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_norm.weight': tensor([ 9.7275e-04, 1.0315e-02, 3.7107e-10, 2.0752e-03, 2.0294e-03, + 2.3723e-05, -1.9550e-05, 5.0964e-03, 3.4180e-02, 2.6001e-02, + 6.8918e-07, 2.4536e-02, 5.5469e-01, 7.4768e-03, 8.4229e-03, + 5.3711e-03, 4.6875e-01, 1.6689e-05, 3.6163e-03, 1.1265e-05, + 5.1498e-04, 4.2725e-03, 5.9814e-02, 1.5991e-02, 1.1963e-02, + 6.3171e-03, 2.5146e-02, -7.5817e-05, 9.9487e-03, 6.7444e-03, + 4.1602e-01, 2.5146e-02, 4.8256e-04, 4.2188e-01, 5.3125e-01, + 1.9684e-03, 3.4790e-03, 3.4668e-02, 1.1475e-01, 1.4893e-02, + 3.8818e-02, 4.0894e-03, 6.4028e-09, 1.4496e-03, 1.8921e-02, + 2.6855e-02, 1.1719e-02, 9.7656e-03, 1.4954e-03, 9.8348e-06, + 8.5449e-03, -9.7156e-06, -7.1526e-06, 1.6809e-05, -4.6539e-04, + -2.6941e-05, 1.8677e-02, 1.0620e-02, 1.0620e-02, 2.9907e-03, + 4.0054e-04, 1.8716e-05, 8.7280e-03, -1.3530e-05, 1.3828e-04, + 1.4782e-04, 1.7047e-05, 2.5391e-01, 7.4387e-04, 1.7090e-02, + 4.0430e-01, -1.9455e-03, 3.9307e-02, 4.9744e-03, 5.4240e-06, + 1.3351e-04, 9.6436e-03, 4.9591e-04, 2.9449e-03, 1.8188e-02, + 2.0386e-02, 1.6968e-02, 1.9043e-01, 6.1951e-03, 3.4094e-05, + 1.5564e-02, 5.4321e-03, 2.0862e-05, 6.1512e-05, 4.7656e-01, + 7.6599e-03, -8.3447e-05, 1.0193e-02, 3.0100e-06, -5.2691e-05, + 4.3555e-01, 2.8419e-04, 7.9346e-03, -1.8304e-11, 1.0834e-03, + 4.1797e-01, 1.7166e-05, 9.5825e-03, -4.8578e-06, 8.3618e-03, + -3.4571e-05, 2.0409e-04, -2.4033e-04, 2.3804e-02, 1.1597e-02, + 6.1951e-03, -4.7982e-06, 3.1494e-02, 1.1301e-04, 1.1841e-02, + 1.4114e-04, 2.5513e-02, 4.1504e-03, 6.7749e-03, 4.1199e-04, + 2.5272e-05, 1.1169e-02, 8.7280e-03, 6.9885e-03, 1.5198e-02, + 2.6512e-04, 2.6855e-03, 5.3167e-05], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_proj.weight': tensor([[-0.0008, 0.0007, 0.0026, ..., -0.0004, -0.0014, 0.0005], + [-0.0063, -0.0021, -0.0051, ..., -0.0028, -0.0053, -0.0034], + [-0.0005, -0.0015, -0.0016, ..., -0.0027, -0.0029, -0.0014], + ..., + [ 0.0026, -0.0025, 0.0018, ..., 0.0011, 0.0024, -0.0006], + [ 0.0010, 0.0005, -0.0018, ..., -0.0025, 0.0009, 0.0019], + [ 0.0035, 0.0019, -0.0023, ..., -0.0041, -0.0006, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_norm.weight': tensor([-9.7275e-04, 1.0254e-02, 3.6016e-10, 1.8921e-03, 2.0142e-03, + 2.7061e-05, 1.9550e-05, 5.0659e-03, 3.3203e-02, 2.6123e-02, + -7.1898e-07, 2.4902e-02, 5.5859e-01, 7.4463e-03, 7.9346e-03, + 5.3406e-03, 4.6875e-01, 1.7524e-05, 3.5248e-03, -2.5392e-05, + 5.7983e-04, 4.3030e-03, 5.9814e-02, 1.5991e-02, 1.1780e-02, + 6.3782e-03, 2.5024e-02, -7.5817e-05, 9.3384e-03, 6.7444e-03, + 4.1602e-01, 2.5024e-02, 4.5395e-04, 4.2188e-01, 5.3125e-01, + 1.8234e-03, 3.5553e-03, 3.4912e-02, 1.1572e-01, 1.4893e-02, + 3.8818e-02, 4.1199e-03, -6.2282e-09, 1.2131e-03, 1.8555e-02, + 2.6733e-02, 1.1719e-02, 9.7656e-03, 1.3580e-03, 9.8944e-06, + 8.4839e-03, -1.0431e-05, -7.1526e-06, 1.9789e-05, 4.6539e-04, + 1.3947e-05, 1.8433e-02, 1.0742e-02, 1.0925e-02, 2.9907e-03, + -2.8419e-04, 1.8716e-05, 8.4229e-03, -1.3888e-05, 8.6308e-05, + 1.8692e-04, 1.6928e-05, 2.5391e-01, 6.4468e-04, 1.7090e-02, + 4.0430e-01, -1.9455e-03, 3.7842e-02, 4.9438e-03, -1.7956e-06, + 1.6117e-04, 8.6670e-03, 4.7112e-04, 2.9297e-03, 1.8188e-02, + 2.0264e-02, 1.7090e-02, 1.9043e-01, 6.0425e-03, 2.6345e-05, + 1.5564e-02, 5.8289e-03, 2.2173e-05, 5.7459e-05, 4.7656e-01, + 7.6599e-03, -9.9182e-05, 9.8877e-03, 2.2948e-06, -5.4359e-05, + 4.3555e-01, -1.0967e-04, 7.9956e-03, 1.8304e-11, 1.0605e-03, + 4.1797e-01, 1.6928e-05, 1.0498e-02, -7.1526e-06, 8.3618e-03, + -3.5048e-05, 2.2697e-04, -2.2984e-04, 2.3438e-02, 1.1536e-02, + 6.2866e-03, -4.2617e-06, 3.1128e-02, 1.1396e-04, 1.2024e-02, + -1.0920e-04, 2.5513e-02, 4.2725e-03, 7.2021e-03, 4.2534e-04, + 2.5153e-05, 1.1292e-02, 8.7280e-03, 6.9885e-03, 1.4771e-02, + 1.6975e-04, 2.7466e-03, -4.8637e-05], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.v_proj.weight': tensor([[-0.0104, 0.0016, 0.0097, ..., 0.0032, 0.0028, -0.0110], + [-0.0243, -0.0013, 0.0063, ..., -0.0010, -0.0076, 0.0008], + [-0.0132, 0.0032, 0.0044, ..., 0.0031, 0.0153, -0.0128], + ..., + [ 0.0117, 0.0008, 0.0018, ..., -0.0023, 0.0064, 0.0181], + [-0.0046, -0.0002, 0.0117, ..., -0.0018, -0.0161, -0.0034], + [-0.0216, -0.0033, 0.0044, ..., -0.0035, -0.0064, -0.0221]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.output_proj.weight': tensor([[ 0.0051, -0.0060, -0.0038, ..., -0.0053, -0.0006, 0.0098], + [ 0.0043, -0.0039, -0.0088, ..., 0.0214, -0.0043, -0.0114], + [-0.0044, 0.0049, 0.0037, ..., 0.0083, 0.0037, 0.0073], + ..., + [-0.0089, -0.0048, 0.0079, ..., -0.0076, -0.0035, -0.0005], + [ 0.0046, -0.0064, 0.0063, ..., 0.0055, -0.0054, -0.0113], + [ 0.0150, -0.0101, -0.0053, ..., 0.0051, -0.0061, 0.0074]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_img.weight': tensor([[ 0.0053, 0.0117, 0.0039, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0072, 0.0249, 0.0291], + [-0.0095, -0.0042, -0.0115, ..., -0.0342, -0.0327, 0.0066], + ..., + [ 0.0078, 0.0045, 0.0045, ..., 0.0086, 0.0286, -0.0165], + [ 0.0068, -0.0053, -0.0101, ..., -0.0101, 0.0142, -0.0309], + [ 0.0242, -0.0102, 0.0255, ..., -0.0227, -0.0044, -0.0153]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, 0.0067], + [ 0.0042, 0.0121, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0084, ..., -0.0059, -0.0057, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, -0.0037, 0.0049], + [-0.0160, 0.0051, 0.0039, ..., 0.0133, -0.0173, -0.0046], + [-0.0051, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0098]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0048, -0.0364, ..., -0.0225, -0.0142, -0.0056], + [-0.0157, 0.0051, -0.0364, ..., 0.0101, 0.0120, 0.0036], + [ 0.0043, 0.0615, 0.0010, ..., -0.0205, -0.0306, -0.0115], + ..., + [-0.0038, -0.0038, -0.0103, ..., -0.0211, -0.0019, 0.0288], + [ 0.0266, -0.0123, -0.0044, ..., -0.0420, -0.0282, 0.0053], + [ 0.0286, 0.0325, 0.0093, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.0.mlp.layer1.weight': tensor([[ 0.0073, 0.0172, -0.0126, ..., 0.0037, -0.0081, 0.0063], + [-0.0317, 0.0056, -0.0040, ..., -0.0054, 0.0044, -0.0088], + [-0.0048, -0.0020, 0.0048, ..., 0.0035, 0.0062, -0.0051], + ..., + [-0.0073, -0.0065, -0.0038, ..., -0.0095, 0.0120, 0.0034], + [ 0.0161, 0.0081, 0.0046, ..., -0.0063, 0.0014, 0.0034], + [-0.0178, -0.0041, -0.0038, ..., 0.0040, -0.0090, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.mlp.layer2.weight': tensor([[ 0.0001, -0.0159, 0.0052, ..., -0.0053, 0.0208, -0.0045], + [-0.0054, -0.0058, -0.0045, ..., -0.0073, 0.0044, -0.0053], + [ 0.0075, 0.0032, 0.0086, ..., 0.0044, 0.0024, -0.0157], + ..., + [ 0.0098, -0.0069, -0.0067, ..., 0.0011, 0.0050, -0.0056], + [-0.0063, 0.0007, 0.0046, ..., 0.0077, 0.0239, -0.0121], + [ 0.0059, -0.0151, 0.0050, ..., 0.0119, -0.0114, -0.0085]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0576, -0.0562, -0.0547, ..., 0.0029, 0.0030, 0.0030], + [ 0.0559, 0.0549, 0.0532, ..., -0.0023, -0.0024, -0.0025], + [ 0.0554, 0.0544, 0.0532, ..., -0.0025, -0.0025, -0.0026], + ..., + [ 0.0566, 0.0557, 0.0535, ..., -0.0022, -0.0024, -0.0024], + [ 0.0535, 0.0525, 0.0505, ..., -0.0021, -0.0021, -0.0021], + [ 0.0571, 0.0559, 0.0547, ..., -0.0029, -0.0030, -0.0029]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_self_attn.2.weight': tensor([[ 0.0019, -0.0016, -0.0020, ..., -0.0028, -0.0024, -0.0018], + [ 0.0017, -0.0021, -0.0018, ..., 0.0024, -0.0017, -0.0018], + [ 0.0002, 0.0033, 0.0002, ..., 0.0035, 0.0034, -0.0003], + ..., + [-0.0374, 0.0452, 0.0442, ..., 0.0420, 0.0374, 0.0374], + [-0.0100, 0.0103, 0.0104, ..., 0.0090, 0.0061, 0.0103], + [ 0.0078, -0.0095, -0.0092, ..., -0.0069, -0.0066, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_cross_attn.1.weight': tensor([[-9.0942e-03, -8.6670e-03, 7.8125e-03, ..., -6.0654e-04, + -5.3406e-04, -7.0572e-04], + [-4.1992e-02, -4.0527e-02, -4.0039e-02, ..., 2.9945e-04, + 3.0518e-04, 3.1471e-04], + [ 4.3213e-02, 4.2236e-02, 4.1748e-02, ..., -8.7738e-05, + -6.7711e-05, -9.2506e-05], + ..., + [ 4.1748e-02, 4.0771e-02, 4.0283e-02, ..., -3.7766e-04, + -3.1090e-04, -3.4714e-04], + [ 5.4688e-02, 5.3223e-02, 5.2002e-02, ..., -2.0981e-04, + -1.3924e-04, -1.9932e-04], + [ 3.8574e-02, 3.6865e-02, 3.6377e-02, ..., -4.7386e-06, + 8.0585e-05, 3.0994e-05]], dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0184, -0.0074, 0.0074, ..., 0.0074, 0.0109, 0.0067], + [-0.0044, -0.0042, 0.0042, ..., 0.0042, 0.0062, 0.0042], + [-0.0007, 0.0049, -0.0077, ..., -0.0049, -0.0074, -0.0078], + ..., + [-0.0078, -0.0265, 0.0332, ..., 0.0265, 0.0432, 0.0286], + [ 0.0114, 0.0063, 0.0122, ..., -0.0059, 0.0040, 0.0134], + [-0.0136, 0.0040, -0.0077, ..., -0.0040, -0.0046, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_mlp.1.weight': tensor([[-0.0037, -0.0033, -0.0031, ..., 0.0002, 0.0004, 0.0004], + [ 0.0540, 0.0518, 0.0503, ..., -0.0020, -0.0023, -0.0024], + [ 0.0459, 0.0444, 0.0437, ..., -0.0033, -0.0036, -0.0036], + ..., + [ 0.0884, 0.0854, 0.0845, ..., -0.0005, -0.0007, -0.0008], + [ 0.0035, 0.0035, 0.0036, ..., -0.0002, -0.0003, -0.0002], + [ 0.0654, 0.0635, 0.0618, ..., -0.0010, -0.0010, -0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.adaln_modulation_mlp.2.weight': tensor([[-0.0004, 0.0125, 0.0136, ..., 0.0120, 0.0003, 0.0098], + [-0.0039, -0.0009, 0.0010, ..., -0.0024, 0.0039, -0.0014], + [-0.0015, 0.0054, 0.0057, ..., 0.0036, 0.0015, 0.0044], + ..., + [ 0.0084, 0.0654, 0.0845, ..., 0.0688, -0.0084, 0.0500], + [ 0.0065, -0.0017, -0.0024, ..., -0.0022, -0.0064, -0.0020], + [-0.0120, -0.0071, -0.0072, ..., -0.0064, 0.0119, -0.0076]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.before_proj.weight': tensor([[ 0.0081, 0.0043, -0.0024, ..., -0.0027, -0.0028, 0.0038], + [-0.0078, -0.0134, -0.0053, ..., 0.0205, 0.0040, 0.0084], + [-0.0042, -0.0051, 0.0022, ..., 0.0023, 0.0043, -0.0065], + ..., + [-0.0031, 0.0089, 0.0085, ..., 0.0024, -0.0022, -0.0054], + [ 0.0143, 0.0198, 0.0146, ..., 0.0020, 0.0107, 0.0045], + [-0.0003, -0.0062, 0.0044, ..., 0.0002, -0.0022, -0.0056]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.before_proj.bias': tensor([ 0.0018, 0.0023, 0.0036, ..., 0.0046, 0.0033, -0.0115], + dtype=torch.bfloat16), 'net.control_blocks_2.0.after_proj.weight': tensor([[-0.0176, 0.0060, -0.0047, ..., -0.0038, -0.0049, -0.0137], + [-0.0106, -0.0037, -0.0068, ..., -0.0107, 0.0072, 0.0188], + [ 0.0019, 0.0009, -0.0058, ..., -0.0017, 0.0006, -0.0014], + ..., + [-0.0062, -0.0015, -0.0006, ..., 0.0024, 0.0011, -0.0017], + [ 0.0145, -0.0029, 0.0059, ..., -0.0047, -0.0109, 0.0077], + [ 0.0077, 0.0045, -0.0050, ..., 0.0039, -0.0055, 0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_2.0.after_proj.bias': tensor([ 0.0300, -0.0136, 0.0002, ..., -0.0012, -0.0262, 0.0337], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_proj.weight': tensor([[-0.0199, 0.0188, 0.0041, ..., 0.0259, -0.0085, 0.0050], + [-0.0106, -0.0162, 0.0114, ..., -0.0073, -0.0008, -0.0393], + [ 0.0139, 0.0078, 0.0051, ..., 0.0129, 0.0022, -0.0157], + ..., + [ 0.0159, 0.0104, -0.0030, ..., 0.0101, -0.0117, -0.0049], + [-0.0114, -0.0149, -0.0050, ..., -0.0016, -0.0167, -0.0042], + [-0.0027, -0.0081, -0.0071, ..., -0.0084, 0.0095, 0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_norm.weight': tensor([ 3.0664e-01, 9.6875e-01, 5.7129e-02, 7.6562e-01, 9.4922e-01, + -6.4319e-09, -3.0640e-02, 5.4123e-15, -4.0149e-19, 5.0093e-13, + 3.9916e-28, -8.3778e-33, -5.4919e-35, 3.3253e-22, -2.3074e-29, + 3.4907e-34, 1.4941e-01, 2.9325e-05, 1.5068e-04, -1.8921e-02, + 5.1880e-03, -1.3447e-04, 2.6367e-01, 9.4922e-01, 1.1875e+00, + 3.4766e-01, 7.3547e-03, 4.2383e-01, 1.7090e-02, 4.6327e-12, + 1.3281e-17, 4.6629e-15, 5.6839e-04, 1.4343e-03, -3.5095e-04, + 1.3184e-02, 1.4782e-04, 2.6375e-10, -3.8338e-04, 4.2848e-16, + 2.6455e-17, -2.0539e-14, -3.1896e-21, 8.8281e-01, 2.7148e-01, + 2.0117e-01, 4.7852e-01, 1.1406e+00, 1.2402e-01, -3.6926e-03, + -2.1118e-02, 2.0020e-01, 2.8992e-04, 1.8917e-10, -3.8147e-06, + 7.4593e-16, 7.2632e-03, 4.1034e-13, 7.5731e-29, -6.0486e-34, + 3.8556e-35, 1.6515e-15, -1.2494e-20, 1.0532e-33, 3.4766e-01, + 2.5586e-01, 5.8203e-01, 2.0996e-01, 1.1914e-01, -3.4925e-09, + 1.9287e-02, 2.2898e-15, -1.7364e-19, -5.2935e-13, -7.8097e-29, + 8.7269e-34, 2.7836e-35, -9.9262e-23, -1.1107e-27, -1.1059e-34, + 1.4191e-03, 2.7466e-04, 1.6403e-04, -1.2970e-04, 3.0708e-04, + 4.2236e-02, 9.4922e-01, 2.1582e-01, 2.0312e-01, 1.0781e+00, + 1.9824e-01, 1.8945e-01, -4.1504e-03, -4.0075e-12, -3.6592e-18, + 3.8303e-15, 1.0757e-03, 1.6724e-02, -3.8910e-03, -4.3640e-03, + 3.3188e-04, 5.5252e-11, 2.2799e-06, -7.3726e-18, 6.6058e-15, + -5.0848e-14, -5.5318e-24, 2.0312e-01, 1.1484e+00, 1.1953e+00, + 1.3750e+00, 4.0234e-01, 3.8281e-01, 9.8877e-03, -2.1484e-02, + -2.2583e-02, -7.4863e-05, 1.9554e-10, 2.4915e-05, 1.3791e-16, + 1.6846e-02, 1.0747e-13, 2.2404e-28, -6.4398e-34, -1.0391e-35, + -1.3214e-19, 7.1998e-20, 7.7037e-34], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.self_attn.k_proj.weight': tensor([[ 0.0152, 0.0017, -0.0059, ..., -0.0053, 0.0058, -0.0119], + [ 0.0231, 0.0071, 0.0044, ..., -0.0135, -0.0038, -0.0149], + [ 0.0060, 0.0164, -0.0024, ..., -0.0013, 0.0027, 0.0041], + ..., + [-0.0061, 0.0035, -0.0077, ..., 0.0115, 0.0005, -0.0031], + [-0.0100, -0.0090, -0.0032, ..., 0.0115, -0.0021, -0.0073], + [-0.0007, -0.0018, -0.0008, ..., -0.0059, 0.0046, 0.0048]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.k_norm.weight': tensor([ 3.2617e-01, 9.8047e-01, -1.0205e-01, 7.8516e-01, 9.4922e-01, + 2.9686e-09, 2.6245e-02, -5.9119e-15, -3.9980e-19, 5.4357e-13, + 4.5438e-28, -3.2259e-33, 4.4387e-35, 3.3418e-22, -2.1201e-29, + 3.4155e-34, 1.4941e-01, 2.9206e-05, -7.9155e-05, 5.4932e-02, + 5.1880e-03, 8.8215e-05, 2.9492e-01, 9.5312e-01, 1.1719e+00, + 1.4258e-01, 1.4844e-01, 4.6289e-01, 1.4404e-02, -4.5759e-12, + 1.3390e-17, 4.4964e-15, 5.3406e-04, -1.9302e-03, -3.2806e-04, + 4.2236e-02, 1.4687e-04, 2.1009e-10, -2.3651e-04, -4.2848e-16, + 2.4395e-17, -1.9540e-14, -3.1896e-21, 8.7891e-01, 2.5586e-01, + -4.7302e-03, -3.2227e-01, 1.1484e+00, 1.3672e-01, -3.0518e-03, + 2.8809e-02, 1.2402e-01, 1.0633e-04, -1.5280e-10, -3.4720e-06, + -7.4593e-16, 8.4229e-03, 4.1034e-13, 6.9814e-29, 9.3287e-35, + -1.2413e-35, -1.6515e-15, -1.1911e-20, -1.9410e-34, 3.8281e-01, + 3.1250e-01, 5.8594e-01, 1.6895e-01, -9.9609e-02, -4.1036e-09, + -2.2827e-02, -1.5543e-15, 4.5401e-19, -5.7909e-13, 2.1299e-29, + 1.0171e-33, 1.9184e-35, 8.2718e-23, 1.0539e-27, -4.8750e-34, + 3.4332e-04, 3.2663e-05, -1.6594e-04, -1.0920e-04, 3.0899e-04, + 4.2236e-02, 9.4531e-01, 2.0898e-01, 1.9824e-01, 1.0703e+00, + 1.8555e-01, 2.2266e-01, 9.3994e-03, 3.8085e-12, -4.4181e-18, + -3.6915e-15, 1.2054e-03, -1.5747e-02, -3.8910e-03, 3.7598e-02, + 3.2997e-04, 2.8922e-10, 1.6317e-06, -4.7976e-18, 6.6891e-15, + -5.0848e-14, 5.8678e-24, 2.9297e-01, 1.1406e+00, 1.1953e+00, + 1.3750e+00, 1.2891e-01, 3.9648e-01, -9.7656e-03, 4.3030e-03, + -1.6992e-01, 3.4332e-04, -1.7280e-10, -2.5034e-05, -1.3618e-16, + -1.5991e-02, -1.0569e-13, 1.9958e-28, -4.8750e-34, -3.7616e-35, + 3.3475e-18, 7.1574e-20, -4.5139e-34], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.self_attn.v_proj.weight': tensor([[ 8.2397e-03, 5.5847e-03, 7.1106e-03, ..., -2.0630e-02, + 8.9722e-03, 4.8828e-03], + [ 1.1108e-02, 6.4697e-03, 8.4686e-04, ..., -3.6163e-03, + -4.8523e-03, -4.3335e-03], + [-2.0905e-03, -3.7079e-03, 9.7275e-05, ..., -6.8359e-03, + 4.9744e-03, -7.9346e-03], + ..., + [ 3.6926e-03, 8.4229e-03, 1.3428e-02, ..., 2.0508e-02, + 3.6926e-03, -1.7578e-02], + [-1.8555e-02, 5.4321e-03, 1.1169e-02, ..., 3.5889e-02, + -7.6294e-03, 6.4850e-04], + [-5.1880e-03, -4.0588e-03, -5.6152e-03, ..., 1.9531e-02, + -5.6458e-03, 4.4861e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.1.self_attn.output_proj.weight': tensor([[-0.0062, -0.0013, -0.0136, ..., 0.0078, -0.0044, -0.0070], + [ 0.0066, 0.0037, 0.0198, ..., 0.0023, -0.0028, 0.0106], + [-0.0044, -0.0047, -0.0019, ..., -0.0060, 0.0150, 0.0024], + ..., + [ 0.0079, 0.0010, -0.0066, ..., -0.0031, -0.0305, -0.0067], + [ 0.0038, -0.0014, -0.0115, ..., -0.0028, -0.0035, -0.0024], + [-0.0033, 0.0103, -0.0242, ..., 0.0129, 0.0017, -0.0036]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_proj.weight': tensor([[ 0.0050, 0.0261, 0.0086, ..., 0.0053, 0.0153, -0.0013], + [ 0.0194, 0.0076, 0.0065, ..., 0.0080, -0.0079, -0.0320], + [ 0.0176, 0.0128, 0.0025, ..., -0.0038, -0.0118, -0.0253], + ..., + [ 0.0061, 0.0032, -0.0038, ..., 0.0105, -0.0020, 0.0168], + [ 0.0058, 0.0090, 0.0104, ..., -0.0047, -0.0013, 0.0248], + [ 0.0175, -0.0170, -0.0084, ..., 0.0095, 0.0043, -0.0243]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_norm.weight': tensor([ 1.8799e-02, 4.7070e-01, 2.3560e-02, 1.0315e-02, 1.2107e-07, + 6.2937e-10, 3.8338e-04, 5.0391e-01, 8.6212e-04, 4.1602e-01, + 1.9165e-02, 1.9287e-02, 3.5095e-03, 1.3000e-02, 2.3315e-02, + 1.3977e-02, 2.6733e-02, -3.8147e-05, -4.0293e-05, 3.5400e-02, + 2.6367e-01, 3.2227e-02, 1.7700e-03, 1.4343e-02, 3.5400e-03, + 4.1008e-05, 1.8716e-05, 4.5410e-02, -6.2883e-06, 3.8281e-01, + 8.7738e-05, -1.3828e-04, 6.3171e-03, -2.5034e-05, 1.0864e-02, + 1.0193e-02, -2.0646e-10, 4.1406e-01, 1.0803e-02, 2.5635e-03, + 7.7148e-02, 2.8534e-03, 9.8877e-03, 1.0490e-04, 7.6904e-03, + 1.3977e-02, 2.9144e-03, 1.7212e-02, 1.4587e-02, 1.4709e-02, + 7.6904e-03, 4.0039e-01, 1.3611e-02, 1.1719e-02, 3.1494e-02, + 4.5538e-05, 9.1309e-02, 1.4709e-02, 1.5564e-02, 1.3672e-02, + 5.4836e-06, 7.6904e-03, 1.1597e-02, -1.4019e-04, 1.1414e-02, + 5.0659e-03, -8.0585e-05, 3.7689e-03, 1.8921e-02, 8.1543e-02, + 5.9128e-04, 6.1646e-03, 5.2979e-02, 4.1748e-02, 2.3956e-03, + 1.2695e-02, 4.0531e-06, 1.8677e-02, 3.1006e-02, 2.0996e-02, + 5.1880e-03, 1.6113e-02, 2.8320e-02, 3.6955e-06, 2.5391e-02, + 9.3994e-03, 3.5400e-03, 3.3203e-02, 6.9275e-03, 2.9602e-03, + 1.6724e-02, 2.0981e-04, 4.2419e-03, 2.1362e-02, 1.6724e-02, + -1.3351e-04, 9.9487e-03, 3.2227e-02, 1.5320e-02, 3.8086e-02, + 3.0762e-02, 2.4780e-02, 1.1169e-02, 6.7711e-05, -4.9360e-08, + 6.4392e-03, 1.8692e-04, 2.0264e-02, 8.1787e-03, 1.2573e-02, + 4.1809e-03, 1.1414e-02, 2.0508e-02, 5.1172e-01, 2.0447e-03, + 9.6436e-03, 1.0376e-02, 5.7068e-03, 7.8735e-03, 7.9590e-02, + 9.2983e-06, 4.5538e-05, 1.6451e-05, 1.2695e-02, 1.5259e-02, + 1.6357e-02, 2.4902e-02, 1.2573e-02], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_proj.weight': tensor([[ 0.0014, -0.0033, -0.0022, ..., -0.0038, 0.0019, -0.0005], + [-0.0085, -0.0035, -0.0021, ..., 0.0098, 0.0134, 0.0070], + [ 0.0055, 0.0036, 0.0045, ..., 0.0072, 0.0049, -0.0034], + ..., + [-0.0057, -0.0049, 0.0025, ..., -0.0029, -0.0009, -0.0032], + [-0.0078, -0.0053, -0.0126, ..., -0.0014, 0.0068, 0.0006], + [-0.0042, -0.0004, -0.0013, ..., 0.0044, -0.0016, -0.0033]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_norm.weight': tensor([ 1.9531e-02, 4.7070e-01, 2.3560e-02, 1.0437e-02, 1.2107e-07, + -2.2828e-10, 4.1962e-04, 5.0391e-01, 8.6594e-04, 4.1602e-01, + 2.0020e-02, 1.9287e-02, 3.5858e-03, 1.3000e-02, 2.3193e-02, + 1.3977e-02, 2.7100e-02, -3.8385e-05, -3.9816e-05, 3.5645e-02, + 2.6367e-01, 3.1982e-02, 1.7776e-03, 1.4587e-02, 3.6163e-03, + 4.0054e-05, 1.8597e-05, 4.5410e-02, -3.4124e-06, 3.8281e-01, + 8.7261e-05, -1.3924e-04, 6.1646e-03, -2.4796e-05, 1.0803e-02, + 1.0315e-02, -1.9008e-10, 4.1406e-01, 1.0864e-02, 2.4567e-03, + 7.6172e-02, 2.8839e-03, 9.6436e-03, 1.0633e-04, 7.5989e-03, + 1.3916e-02, 2.8687e-03, 1.7090e-02, 1.4404e-02, 1.5015e-02, + 7.5989e-03, 4.0039e-01, 1.3611e-02, 1.1230e-02, 3.1250e-02, + 4.4346e-05, 9.1309e-02, 1.4771e-02, 1.5259e-02, 1.3794e-02, + 6.3479e-06, 8.0566e-03, 1.1597e-02, -1.4019e-04, 1.1475e-02, + 4.9744e-03, -8.0585e-05, 3.7689e-03, 1.9043e-02, 8.1543e-02, + 5.6839e-04, 6.1340e-03, 5.2979e-02, 4.1748e-02, 2.4261e-03, + 1.2878e-02, 4.9472e-06, 1.8555e-02, 3.0884e-02, 2.1118e-02, + 5.1880e-03, 1.6113e-02, 2.7710e-02, 4.0531e-06, 2.5513e-02, + 9.3994e-03, 3.3875e-03, 3.2959e-02, 7.2632e-03, 3.0365e-03, + 1.6968e-02, 2.1172e-04, 4.3335e-03, 2.1240e-02, 1.6602e-02, + -1.3351e-04, 9.2773e-03, 3.1982e-02, 1.5259e-02, 3.7598e-02, + 3.1128e-02, 2.4780e-02, 1.1230e-02, 6.6280e-05, -1.7229e-08, + 6.4392e-03, 1.9360e-04, 1.9775e-02, 8.3008e-03, 1.2451e-02, + 4.2114e-03, 1.1169e-02, 2.0020e-02, 5.1172e-01, 2.0752e-03, + 9.5215e-03, 1.0315e-02, 5.5847e-03, 7.9956e-03, 7.9590e-02, + 1.3411e-05, 4.4584e-05, 1.6093e-05, 1.3000e-02, 1.5320e-02, + 1.6602e-02, 2.5024e-02, 1.2329e-02], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.v_proj.weight': tensor([[-0.0210, 0.0036, 0.0016, ..., 0.0012, -0.0206, -0.0220], + [-0.0035, 0.0081, 0.0038, ..., 0.0037, 0.0134, 0.0095], + [-0.0422, -0.0024, -0.0076, ..., -0.0079, -0.0286, -0.0074], + ..., + [-0.0376, 0.0016, -0.0150, ..., 0.0014, 0.0220, 0.0054], + [ 0.0082, 0.0015, -0.0062, ..., 0.0018, 0.0128, 0.0084], + [ 0.0112, 0.0050, -0.0072, ..., 0.0006, 0.0069, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.output_proj.weight': tensor([[ 7.2327e-03, 9.3384e-03, 4.6692e-03, ..., -6.8359e-03, + 2.5330e-03, 6.9885e-03], + [-3.9368e-03, -1.3184e-02, 6.2866e-03, ..., -8.1787e-03, + 6.6833e-03, 8.4229e-03], + [ 2.6245e-03, -2.0599e-03, -1.9684e-03, ..., -1.6785e-04, + -6.1417e-04, -7.0190e-03], + ..., + [-6.5918e-03, 6.3896e-05, 4.1809e-03, ..., 5.4626e-03, + 2.7466e-03, 7.8735e-03], + [ 7.6294e-03, 1.1658e-02, -5.0735e-04, ..., -2.6123e-02, + 7.1106e-03, 2.1606e-02], + [ 4.5166e-03, 4.0588e-03, -1.7456e-02, ..., -1.0742e-02, + 6.0120e-03, 7.8735e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_img.weight': tensor([[ 0.0079, 0.0034, 0.0430, ..., -0.0327, 0.0042, 0.0146], + [-0.0278, 0.0280, 0.0249, ..., -0.0046, -0.0378, 0.0271], + [ 0.0142, -0.0120, 0.0312, ..., -0.0074, 0.0253, -0.0118], + ..., + [ 0.0245, -0.0152, 0.0027, ..., -0.0452, 0.0216, -0.0209], + [-0.0071, 0.0074, -0.0471, ..., -0.0101, 0.0157, -0.0083], + [ 0.0085, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0028, -0.0122, -0.0115], + [ 0.0137, 0.0062, -0.0188, ..., 0.0157, 0.0082, -0.0047], + [-0.0306, -0.0051, -0.0273, ..., 0.0103, 0.0337, -0.0048], + ..., + [-0.0071, 0.0003, 0.0063, ..., 0.0209, -0.0106, 0.0081], + [ 0.0194, 0.0210, 0.0123, ..., -0.0146, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0109, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img.weight': tensor([[-0.0003, -0.0076, 0.0173, ..., 0.0084, 0.0356, 0.0217], + [-0.0498, -0.0049, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0104, -0.0013, -0.0139, ..., 0.0041, 0.0269, 0.0082], + ..., + [-0.0045, 0.0120, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0077, -0.0162, ..., -0.0059, -0.0076, -0.0121], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0134, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.1.mlp.layer1.weight': tensor([[-0.0031, -0.0219, -0.0079, ..., 0.0020, -0.0061, -0.0039], + [ 0.0006, 0.0044, 0.0055, ..., 0.0060, -0.0048, -0.0092], + [-0.0221, 0.0100, 0.0028, ..., -0.0045, -0.0090, -0.0073], + ..., + [ 0.0286, 0.0098, 0.0014, ..., 0.0140, 0.0062, 0.0143], + [-0.0159, -0.0058, 0.0041, ..., 0.0157, -0.0101, -0.0102], + [ 0.0052, -0.0114, -0.0042, ..., -0.0054, 0.0070, -0.0181]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.mlp.layer2.weight': tensor([[-0.0096, -0.0122, -0.0084, ..., -0.0148, -0.0040, -0.0079], + [ 0.0123, 0.0056, 0.0131, ..., 0.0037, 0.0097, -0.0077], + [ 0.0052, -0.0040, -0.0037, ..., -0.0010, -0.0015, 0.0038], + ..., + [ 0.0037, -0.0034, 0.0304, ..., -0.0075, -0.0130, 0.0051], + [ 0.0050, 0.0050, 0.0135, ..., 0.0075, 0.0049, -0.0062], + [ 0.0220, -0.0164, -0.0082, ..., 0.0068, 0.0042, 0.0048]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0432, -0.0420, -0.0415, ..., 0.0050, 0.0053, 0.0052], + [ 0.0352, 0.0339, 0.0330, ..., -0.0036, -0.0035, -0.0036], + [-0.0422, -0.0403, -0.0398, ..., 0.0050, 0.0051, 0.0050], + ..., + [-0.0476, -0.0464, -0.0449, ..., 0.0034, 0.0035, 0.0036], + [-0.0413, -0.0410, -0.0393, ..., 0.0036, 0.0039, 0.0039], + [-0.0251, -0.0240, -0.0228, ..., 0.0011, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0042, -0.0036, 0.0035, ..., 0.0033, 0.0033, 0.0034], + [ 0.0006, 0.0030, 0.0020, ..., 0.0028, 0.0019, -0.0021], + [-0.0068, 0.0038, -0.0041, ..., -0.0077, -0.0069, -0.0038], + ..., + [-0.0231, -0.0007, -0.0286, ..., -0.0308, -0.0292, 0.0007], + [-0.0117, -0.0054, -0.0129, ..., -0.0142, -0.0121, 0.0092], + [ 0.0069, -0.0063, 0.0044, ..., 0.0069, 0.0062, -0.0011]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_cross_attn.1.weight': tensor([[ 6.2500e-02, 6.0791e-02, 6.0303e-02, ..., -1.9550e-05, + -4.8876e-05, -4.3213e-06], + [ 7.8125e-03, 7.8125e-03, 7.8125e-03, ..., -3.9577e-05, + -1.1349e-04, -2.2888e-04], + [-2.0264e-02, -1.9531e-02, -1.8555e-02, ..., -6.7139e-04, + -6.7139e-04, -6.8283e-04], + ..., + [-2.6489e-02, -2.6489e-02, -2.5635e-02, ..., -3.9864e-04, + -3.5667e-04, -3.1853e-04], + [-2.7100e-02, -2.6489e-02, -2.5635e-02, ..., -2.0123e-04, + -1.5545e-04, -2.0885e-04], + [ 5.7373e-02, 5.5664e-02, 5.4932e-02, ..., -2.0409e-04, + -8.5831e-05, -2.3174e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0074, -0.0041, -0.0145, ..., 0.0043, 0.0080, -0.0095], + [ 0.0068, 0.0026, -0.0005, ..., -0.0011, -0.0033, 0.0017], + [ 0.0147, -0.0042, -0.0042, ..., 0.0043, -0.0070, 0.0095], + ..., + [ 0.0679, 0.0045, -0.0058, ..., 0.0059, -0.0175, 0.0630], + [ 0.0078, -0.0104, -0.0044, ..., -0.0155, -0.0220, 0.0183], + [-0.0065, 0.0087, -0.0153, ..., 0.0106, 0.0091, -0.0083]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_mlp.1.weight': tensor([[ 0.0096, 0.0090, 0.0082, ..., 0.0003, 0.0002, 0.0002], + [ 0.0325, 0.0317, 0.0315, ..., -0.0037, -0.0037, -0.0036], + [ 0.0457, 0.0444, 0.0439, ..., -0.0038, -0.0038, -0.0038], + ..., + [ 0.0085, 0.0082, 0.0088, ..., -0.0020, -0.0020, -0.0019], + [-0.0181, -0.0167, -0.0159, ..., 0.0015, 0.0014, 0.0014], + [ 0.0052, 0.0052, 0.0052, ..., -0.0024, -0.0025, -0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.adaln_modulation_mlp.2.weight': tensor([[-0.0024, -0.0014, -0.0033, ..., 0.0035, 0.0031, 0.0024], + [-0.0004, -0.0010, 0.0036, ..., 0.0014, 0.0003, 0.0015], + [ 0.0036, -0.0046, -0.0056, ..., -0.0038, -0.0011, -0.0038], + ..., + [-0.0011, 0.0083, 0.0131, ..., -0.0092, 0.0016, -0.0082], + [ 0.0167, 0.0129, 0.0238, ..., -0.0120, -0.0175, -0.0125], + [ 0.0157, -0.0153, -0.0156, ..., -0.0150, -0.0161, -0.0140]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.after_proj.weight': tensor([[-0.0284, 0.0079, 0.0044, ..., -0.0072, -0.0132, 0.0128], + [ 0.0086, 0.0190, 0.0042, ..., -0.0061, -0.0064, 0.0031], + [ 0.0120, -0.0156, -0.0128, ..., -0.0083, 0.0104, 0.0078], + ..., + [-0.0187, -0.0171, 0.0166, ..., -0.0065, -0.0040, -0.0048], + [-0.0098, -0.0085, 0.0060, ..., 0.0056, 0.0134, -0.0042], + [-0.0182, 0.0089, -0.0125, ..., -0.0096, -0.0066, -0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_2.1.after_proj.bias': tensor([ 0.0125, 0.0034, 0.0019, ..., 0.0067, 0.0037, -0.0056], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_proj.weight': tensor([[ 0.0243, -0.0233, -0.0165, ..., 0.0219, -0.0100, -0.0254], + [-0.0001, -0.0242, -0.0025, ..., -0.0129, 0.0074, 0.0037], + [ 0.0148, 0.0154, 0.0070, ..., 0.0044, -0.0095, 0.0009], + ..., + [-0.0036, -0.0067, -0.0034, ..., 0.0040, 0.0106, -0.0075], + [ 0.0281, 0.0144, -0.0022, ..., 0.0077, -0.0160, 0.0281], + [ 0.0074, -0.0193, 0.0089, ..., -0.0330, 0.0123, -0.0231]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_norm.weight': tensor([ 1.5625e+00, 4.9414e-01, 5.1562e-01, 3.0273e-01, 1.4922e+00, + 4.3335e-03, -1.1719e-02, -4.3392e-05, 1.8616e-03, 8.9722e-03, + 1.4827e-06, 3.5623e-08, 1.2266e+00, -6.4392e-03, -3.1876e-17, + -3.2715e-02, -1.3367e-02, -4.3213e-02, 2.4414e-03, 1.8188e-02, + 1.6895e-01, 2.8372e-05, 6.4062e-01, 1.5859e+00, 6.0547e-01, + 6.8750e-01, 4.7656e-01, 1.6699e-01, 2.0020e-02, 5.5847e-03, + 2.7710e-02, -4.9591e-05, 6.1340e-03, -2.0874e-02, 3.3691e-02, + 4.4434e-02, 1.2451e-01, 8.9722e-03, -7.6599e-03, 1.5736e-05, + -5.0545e-05, -2.2507e-04, 1.2146e-02, 1.3125e+00, 1.3184e-01, + 1.3906e+00, 6.6797e-01, 2.1094e-01, -3.7598e-02, -1.5869e-02, + 2.9297e-02, 5.1025e-02, 1.3477e-01, 3.3855e-05, 2.4080e-05, + -5.6252e-07, 4.5061e-05, -7.8678e-05, -6.1279e-02, -4.2783e-09, + 4.6875e-02, -1.9932e-04, -3.3691e-02, -3.5889e-02, 8.7500e-01, + 1.6016e+00, 1.4688e+00, 1.4531e+00, 3.1250e-01, -4.6997e-03, + 5.7068e-03, 3.9339e-05, 1.2939e-02, -4.2114e-03, -1.9932e-04, + 9.0525e-07, 3.2715e-02, 2.4780e-02, 7.6050e-15, 8.1250e-01, + 1.0078e+00, -4.6387e-03, -1.1368e-03, 3.2959e-02, 1.2988e-01, + 2.0508e-02, 1.5234e+00, 5.6641e-01, 1.3984e+00, 1.5000e+00, + 2.8125e-01, 5.1562e-01, 5.1025e-02, -1.1902e-02, 2.1606e-02, + -1.8024e-04, -2.0752e-03, -3.7231e-03, 1.8463e-03, 3.2959e-02, + 1.2207e-01, 2.5177e-04, 1.1641e+00, 1.7405e-05, 2.4915e-05, + -4.7684e-05, -8.5354e-05, 9.6875e-01, 1.5391e+00, 1.4844e+00, + 1.5469e+00, 3.7891e-01, 3.2227e-02, 4.5410e-02, 2.2827e-02, + 1.9287e-02, 1.4746e-01, 6.7711e-05, -7.6890e-06, 1.7658e-06, + 1.8120e-05, 9.2030e-05, 1.1094e+00, 2.7831e-10, 2.4536e-02, + -1.2219e-05, 7.6562e-01, 1.1172e+00], dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.self_attn.k_proj.weight': tensor([[-0.0388, 0.0049, -0.0071, ..., 0.0249, 0.0139, 0.0238], + [-0.0056, 0.0130, -0.0036, ..., -0.0114, 0.0283, -0.0051], + [-0.0294, 0.0200, -0.0027, ..., 0.0057, 0.0228, -0.0034], + ..., + [-0.0057, 0.0018, -0.0045, ..., 0.0026, 0.0011, 0.0127], + [ 0.0050, 0.0120, -0.0009, ..., -0.0046, -0.0048, 0.0089], + [ 0.0190, -0.0225, -0.0019, ..., -0.0131, 0.0058, -0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.k_norm.weight': tensor([ 1.6328e+00, 8.9844e-01, 3.9648e-01, 3.6328e-01, 1.4922e+00, + 4.3030e-03, 1.6235e-02, -4.3392e-05, 6.6528e-03, 8.9722e-03, + 3.5018e-06, -1.6065e-08, 1.2266e+00, -6.0425e-03, -8.1966e-17, + -2.6489e-02, -4.1199e-03, 1.1768e-01, 4.2725e-03, 1.9836e-03, + 2.7466e-02, -7.7724e-05, 6.5625e-01, 1.4297e+00, 2.5391e-01, + 7.0312e-01, 5.5859e-01, 2.3926e-01, -1.8677e-02, 1.3000e-02, + 2.9663e-02, 1.3947e-05, 6.5308e-03, 5.4688e-02, 3.4912e-02, + -1.8311e-02, -1.2573e-02, 8.9722e-03, 2.6489e-02, 1.5974e-05, + -8.2016e-05, -2.2030e-04, 1.2024e-02, 1.5078e+00, 9.9219e-01, + 1.3750e+00, 7.3047e-01, 2.5586e-01, 6.5308e-03, 6.7871e-02, + 3.1738e-02, 7.4158e-03, 3.6774e-03, 6.2466e-05, -1.4484e-05, + -4.8429e-07, 2.9206e-05, 4.9114e-05, 9.8877e-03, -4.5693e-09, + -2.2217e-02, -1.9741e-04, -4.4434e-02, -3.8818e-02, 7.5781e-01, + 1.4609e+00, 1.5000e+00, 1.4375e+00, 2.9688e-01, 4.4556e-03, + -1.6937e-03, 3.6716e-05, 2.1515e-03, 4.3030e-03, -1.9932e-04, + -8.9034e-07, -3.0029e-02, 2.5024e-02, 7.4940e-15, 8.1250e-01, + 1.0078e+00, 5.7617e-02, 1.6022e-03, 1.1719e-02, -3.0884e-02, + 2.0752e-02, 1.4844e+00, 9.2969e-01, 1.5156e+00, 1.5391e+00, + 4.8584e-02, 5.1172e-01, 9.9487e-03, 6.8970e-03, 1.5625e-02, + 2.0885e-04, -6.2180e-04, 4.5166e-02, 4.4250e-03, 1.0498e-02, + 1.1414e-02, -1.1492e-04, 1.1641e+00, 1.7643e-05, 2.4319e-05, + -4.7207e-05, -4.1246e-05, 6.7188e-01, 1.3672e+00, 1.3828e+00, + 1.6641e+00, 3.8086e-01, -4.3701e-02, 1.1963e-01, 2.4170e-02, + 2.2583e-02, -1.5259e-02, -1.0729e-06, -2.0266e-06, -1.4156e-06, + 4.1485e-05, -1.6212e-04, 1.1094e+00, -3.0923e-10, -1.7853e-03, + -8.9407e-06, 7.6562e-01, 1.1172e+00], dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.self_attn.v_proj.weight': tensor([[ 0.0025, -0.0052, -0.0096, ..., -0.0293, -0.0051, 0.0109], + [ 0.0068, 0.0021, -0.0122, ..., -0.0017, -0.0008, -0.0042], + [-0.0245, 0.0041, -0.0022, ..., 0.0148, 0.0055, -0.0143], + ..., + [ 0.0137, 0.0065, -0.0337, ..., -0.0454, 0.0104, -0.0117], + [ 0.0017, -0.0108, 0.0018, ..., -0.0069, -0.0074, -0.0117], + [-0.0015, 0.0028, -0.0188, ..., -0.0042, 0.0068, 0.0016]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.self_attn.output_proj.weight': tensor([[-0.0040, 0.0315, -0.0271, ..., -0.0048, -0.0188, 0.0134], + [ 0.0038, -0.0175, -0.0282, ..., 0.0054, 0.0064, -0.0032], + [-0.0034, -0.0038, 0.0041, ..., 0.0038, -0.0100, 0.0074], + ..., + [ 0.0094, -0.0041, -0.0128, ..., 0.0277, 0.0123, -0.0049], + [ 0.0029, 0.0190, -0.0044, ..., 0.0171, 0.0071, -0.0211], + [-0.0037, -0.0123, 0.0038, ..., -0.0239, 0.0061, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_proj.weight': tensor([[ 0.0055, -0.0037, 0.0042, ..., -0.0042, 0.0129, 0.0038], + [ 0.0061, 0.0057, 0.0023, ..., 0.0026, 0.0038, -0.0031], + [ 0.0036, -0.0056, 0.0003, ..., -0.0069, 0.0186, -0.0042], + ..., + [-0.0025, 0.0046, 0.0038, ..., 0.0108, 0.0037, 0.0046], + [-0.0059, -0.0049, 0.0035, ..., 0.0124, 0.0095, 0.0079], + [ 0.0080, 0.0013, -0.0031, ..., -0.0109, 0.0103, -0.0040]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_norm.weight': tensor([ 6.7749e-03, 1.2305e-01, 2.2339e-02, 6.0156e-01, 2.3315e-02, + 9.6436e-03, 2.0386e-02, 4.0588e-03, 1.5137e-02, 1.0645e-01, + 2.2583e-02, 4.7852e-02, 2.6172e-01, 2.6489e-02, 1.7969e-01, + 1.8433e-02, 4.6875e-02, 2.4048e-02, 5.5847e-03, 8.3008e-03, + 1.2268e-02, 1.5918e-01, 2.7148e-01, 1.9238e-01, 2.1118e-02, + 1.4941e-01, 4.4922e-02, 8.5449e-03, 4.0820e-01, 1.3611e-02, + 6.3782e-03, 5.0391e-01, 4.4922e-02, 4.1748e-02, 2.0752e-02, + 8.9722e-03, 7.5684e-02, 5.2344e-01, 5.9375e-01, 3.5400e-03, + 1.3062e-02, 1.0156e-01, 8.3618e-03, 5.5859e-01, 1.6479e-02, + 3.2715e-02, 8.4839e-03, 4.2773e-01, 4.2419e-03, 2.1484e-02, + 6.3477e-02, 2.2949e-02, 1.7334e-02, 5.6641e-01, -7.0095e-05, + 3.2227e-02, 3.3691e-02, 6.7383e-02, 1.5198e-02, 1.5198e-02, + 2.5330e-03, 1.2329e-02, 1.5717e-03, 3.5547e-01, 1.4648e-02, + -1.2684e-04, 1.8433e-02, 5.8350e-02, 7.4707e-02, 9.2163e-03, + 9.2773e-03, 6.3672e-01, 1.9789e-05, 6.0938e-01, 1.5381e-02, + 6.7139e-04, 9.5825e-03, 2.9297e-01, 2.9907e-02, 2.9564e-04, + 9.3262e-02, 8.3008e-03, 4.1748e-02, 7.6294e-03, 2.4170e-02, + 5.2490e-03, 2.2827e-02, 6.4941e-02, 2.0898e-01, 9.8877e-03, + 5.9814e-03, 1.3489e-02, 4.0527e-02, 2.0752e-02, 3.5156e-02, + 2.5269e-02, 9.3384e-03, 2.4414e-02, 5.0293e-02, 1.9531e-02, + 2.9907e-02, 1.5625e-02, -3.8300e-03, 5.1514e-02, 6.1719e-01, + 5.2734e-02, 1.1749e-03, 4.9219e-01, 1.8799e-02, 2.3315e-02, + 1.1902e-02, 2.0294e-03, -6.1512e-05, 8.0078e-02, 4.6094e-01, + 1.4099e-02, 2.3682e-02, 6.1328e-01, 1.1084e-01, 3.0518e-02, + 1.1169e-02, 6.9922e-01, 1.4453e-01, 1.3916e-02, 7.8125e-03, + 2.4414e-03, 1.8921e-02, 1.1520e-03], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_proj.weight': tensor([[ 0.0038, -0.0034, 0.0050, ..., -0.0018, -0.0112, 0.0114], + [ 0.0039, 0.0003, -0.0025, ..., 0.0014, -0.0126, -0.0041], + [ 0.0024, -0.0016, -0.0035, ..., 0.0120, -0.0132, -0.0042], + ..., + [ 0.0052, 0.0046, -0.0034, ..., 0.0023, 0.0061, -0.0125], + [ 0.0093, 0.0029, -0.0020, ..., -0.0042, 0.0042, 0.0027], + [-0.0091, 0.0007, -0.0101, ..., -0.0042, -0.0160, 0.0046]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_norm.weight': tensor([ 6.7749e-03, 1.2305e-01, 2.2461e-02, 6.0156e-01, 2.3193e-02, + 9.4604e-03, 2.0386e-02, 4.0588e-03, 1.5381e-02, 1.0645e-01, + 2.2583e-02, 4.6143e-02, 2.6172e-01, 2.6611e-02, 1.7969e-01, + 1.9409e-02, 4.6387e-02, 2.4048e-02, 5.5237e-03, 8.1787e-03, + 1.2390e-02, 1.6016e-01, 2.7344e-01, 1.9238e-01, 2.0752e-02, + 1.4941e-01, 4.3945e-02, 6.3477e-03, 4.1016e-01, 1.3489e-02, + 6.6528e-03, 5.0391e-01, 4.4922e-02, 4.1992e-02, 2.1118e-02, + 8.3008e-03, 7.4707e-02, 5.2344e-01, 5.9375e-01, 3.6011e-03, + 1.3062e-02, 1.0107e-01, 8.3618e-03, 5.5859e-01, 1.6846e-02, + 3.2959e-02, 8.7891e-03, 4.2773e-01, 4.3335e-03, 2.1362e-02, + 6.3477e-02, 2.3193e-02, 1.7334e-02, 5.6641e-01, -7.0095e-05, + 3.2227e-02, 3.3203e-02, 6.7383e-02, 1.5259e-02, 1.3672e-02, + 2.4261e-03, 1.2268e-02, 2.3956e-03, 3.5547e-01, 1.4221e-02, + 2.9755e-04, 1.7822e-02, 5.8350e-02, 7.4707e-02, 9.3384e-03, + 9.5825e-03, 6.3672e-01, 1.9789e-05, 6.0938e-01, 1.5625e-02, + -5.8365e-04, 9.6436e-03, 2.9297e-01, 3.0640e-02, 3.8910e-04, + 9.3750e-02, 8.1177e-03, 4.0771e-02, 7.3547e-03, 2.3682e-02, + 5.2490e-03, 2.3193e-02, 6.5430e-02, 2.0898e-01, 9.8877e-03, + 6.0120e-03, 1.2329e-02, 4.0527e-02, 2.0996e-02, 3.6133e-02, + 2.5269e-02, 9.3994e-03, 2.4414e-02, 5.0293e-02, 1.9409e-02, + 2.9175e-02, 1.5747e-02, 3.8300e-03, 5.2246e-02, 6.1719e-01, + 5.2734e-02, 1.1444e-03, 4.9219e-01, 1.8433e-02, 2.3071e-02, + 1.2695e-02, 2.0294e-03, -5.4836e-05, 8.0078e-02, 4.6094e-01, + 1.3672e-02, 2.3438e-02, 6.1328e-01, 1.1182e-01, 3.0640e-02, + 1.0864e-02, 7.0312e-01, 1.4355e-01, 1.2756e-02, 7.5378e-03, + 2.3346e-03, 1.8677e-02, -9.3460e-04], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.v_proj.weight': tensor([[ 0.0133, -0.0066, -0.0176, ..., -0.0031, -0.0162, -0.0116], + [-0.0157, 0.0014, -0.0052, ..., -0.0026, -0.0210, 0.0210], + [-0.0116, -0.0044, -0.0012, ..., -0.0032, 0.0299, 0.0249], + ..., + [-0.0052, 0.0014, -0.0125, ..., 0.0035, 0.0112, -0.0006], + [ 0.0284, 0.0068, 0.0254, ..., 0.0093, 0.0096, 0.0153], + [ 0.0162, 0.0038, 0.0074, ..., 0.0071, -0.0186, -0.0244]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.output_proj.weight': tensor([[ 0.0153, 0.0042, -0.0161, ..., 0.0133, 0.0081, 0.0150], + [ 0.0088, -0.0010, -0.0029, ..., 0.0084, 0.0030, 0.0031], + [-0.0039, -0.0069, 0.0044, ..., -0.0048, 0.0067, -0.0067], + ..., + [ 0.0047, -0.0082, 0.0043, ..., 0.0047, -0.0096, 0.0056], + [ 0.0045, -0.0133, 0.0067, ..., -0.0403, -0.0249, -0.0233], + [-0.0021, -0.0025, 0.0154, ..., 0.0086, 0.0036, 0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0068, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [ 0.0038, 0.0150, -0.0159, ..., -0.0034, -0.0056, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0068, -0.0057], + [-0.0337, -0.0031, 0.0048, ..., -0.0032, 0.0172, 0.0231], + [-0.0255, 0.0004, -0.0044, ..., 0.0408, 0.0430, -0.0120]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0033], + [ 0.0198, -0.0208, 0.0129, ..., 0.0089, -0.0337, -0.0051], + [ 0.0302, 0.0101, -0.0018, ..., -0.0304, 0.0199, -0.0039], + ..., + [ 0.0171, -0.0276, -0.0066, ..., 0.0231, 0.0182, 0.0011], + [ 0.0221, 0.0388, 0.0047, ..., 0.0042, -0.0488, -0.0361], + [ 0.0045, 0.0118, -0.0059, ..., 0.0010, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img.weight': tensor([[ 0.0273, -0.0041, 0.0118, ..., 0.0022, 0.0347, 0.0615], + [-0.0258, -0.0047, -0.0197, ..., 0.0139, -0.0042, 0.0110], + [-0.0082, -0.0168, -0.0427, ..., -0.0586, 0.0154, 0.0085], + ..., + [ 0.0097, 0.0132, 0.0184, ..., 0.0110, 0.0089, 0.0515], + [-0.0017, -0.0247, -0.0278, ..., -0.0090, -0.0388, 0.0084], + [ 0.0190, -0.0267, -0.0220, ..., -0.0457, -0.0160, 0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.2.mlp.layer1.weight': tensor([[-0.0108, 0.0063, 0.0072, ..., -0.0044, -0.0053, -0.0164], + [-0.0096, 0.0017, -0.0006, ..., -0.0088, 0.0035, -0.0081], + [-0.0038, 0.0062, 0.0038, ..., -0.0022, -0.0123, -0.0510], + ..., + [-0.0118, 0.0111, 0.0038, ..., 0.0068, -0.0043, -0.0075], + [-0.0062, 0.0075, 0.0042, ..., -0.0066, -0.0088, -0.0383], + [-0.0030, 0.0275, -0.0038, ..., 0.0016, 0.0046, -0.0094]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.mlp.layer2.weight': tensor([[-0.0107, -0.0052, -0.0210, ..., 0.0140, -0.0079, -0.0112], + [-0.0051, 0.0047, -0.0121, ..., -0.0032, 0.0078, -0.0229], + [-0.0051, 0.0137, -0.0172, ..., 0.0005, 0.0127, -0.0106], + ..., + [-0.0009, 0.0081, -0.0381, ..., 0.0082, 0.0142, -0.0155], + [-0.0127, 0.0184, 0.0122, ..., 0.0067, 0.0096, -0.0037], + [-0.0101, -0.0040, -0.0364, ..., -0.0003, 0.0109, -0.0334]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0289, 0.0291, 0.0275, ..., -0.0041, -0.0042, -0.0043], + [-0.0254, -0.0244, -0.0242, ..., 0.0043, 0.0047, 0.0046], + [-0.0297, -0.0293, -0.0286, ..., 0.0045, 0.0048, 0.0048], + ..., + [-0.0322, -0.0312, -0.0305, ..., 0.0021, 0.0022, 0.0023], + [-0.0327, -0.0325, -0.0317, ..., 0.0020, 0.0021, 0.0021], + [-0.0046, -0.0046, -0.0045, ..., 0.0001, 0.0001, 0.0001]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0049, -0.0042, -0.0014, ..., -0.0048, -0.0051, -0.0014], + [ 0.0017, -0.0020, -0.0017, ..., -0.0018, -0.0022, 0.0031], + [ 0.0099, -0.0096, -0.0102, ..., -0.0041, -0.0034, 0.0007], + ..., + [ 0.0388, -0.0427, -0.0432, ..., -0.0249, -0.0217, 0.0142], + [-0.0041, 0.0042, -0.0049, ..., 0.0044, 0.0045, 0.0060], + [-0.0148, 0.0173, 0.0148, ..., 0.0162, 0.0146, 0.0165]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_cross_attn.1.weight': tensor([[ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., -1.3065e-04, + -6.0320e-05, -2.1267e-04], + [ 6.1768e-02, 6.0303e-02, 5.8594e-02, ..., 3.3569e-04, + 4.4632e-04, 4.7874e-04], + [-5.3955e-02, -5.2490e-02, -5.1025e-02, ..., -5.2154e-06, + 5.0545e-05, -5.7459e-05], + ..., + [ 3.3447e-02, 3.2715e-02, 3.1982e-02, ..., 1.0204e-04, + 3.3617e-05, 6.2943e-05], + [ 5.9570e-02, 5.8105e-02, 5.7617e-02, ..., 5.2261e-04, + 5.6458e-04, 5.5695e-04], + [-4.8828e-02, -4.8096e-02, -4.7363e-02, ..., -3.2997e-04, + -3.7384e-04, -3.3379e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0254, 0.0063, -0.0156, ..., 0.0030, -0.0043, -0.0016], + [-0.0216, 0.0137, 0.0175, ..., -0.0090, -0.0137, 0.0099], + [ 0.0098, 0.0008, -0.0087, ..., 0.0060, 0.0008, -0.0008], + ..., + [ 0.0225, 0.0156, -0.0128, ..., 0.0072, 0.0080, 0.0045], + [ 0.0117, 0.0444, -0.0057, ..., 0.0040, 0.0225, -0.0079], + [-0.0187, 0.0161, 0.0192, ..., -0.0135, -0.0082, 0.0210]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_mlp.1.weight': tensor([[-0.0571, -0.0557, -0.0542, ..., 0.0029, 0.0031, 0.0031], + [-0.0259, -0.0244, -0.0232, ..., 0.0008, 0.0008, 0.0007], + [ 0.0344, 0.0332, 0.0332, ..., -0.0016, -0.0016, -0.0016], + ..., + [ 0.0304, 0.0294, 0.0295, ..., -0.0018, -0.0017, -0.0015], + [ 0.0356, 0.0342, 0.0327, ..., -0.0010, -0.0009, -0.0010], + [-0.0295, -0.0288, -0.0282, ..., 0.0021, 0.0021, 0.0020]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.adaln_modulation_mlp.2.weight': tensor([[-0.0233, -0.0145, 0.0118, ..., 0.0041, 0.0162, -0.0124], + [-0.0010, -0.0058, -0.0026, ..., -0.0016, 0.0057, 0.0014], + [-0.0004, -0.0009, 0.0005, ..., -0.0008, 0.0005, 0.0001], + ..., + [ 0.0432, -0.0041, -0.0190, ..., -0.0137, 0.0040, 0.0197], + [ 0.0226, -0.0126, -0.0173, ..., -0.0151, 0.0124, 0.0167], + [ 0.0072, -0.0306, -0.0190, ..., -0.0220, 0.0315, 0.0182]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.after_proj.weight': tensor([[-0.0034, 0.0076, -0.0066, ..., -0.0079, 0.0073, 0.0118], + [ 0.0095, -0.0077, 0.0047, ..., -0.0018, 0.0079, 0.0058], + [ 0.0027, 0.0077, 0.0167, ..., 0.0063, 0.0069, 0.0176], + ..., + [ 0.0086, -0.0040, 0.0145, ..., 0.0044, -0.0072, 0.0212], + [ 0.0033, -0.0074, 0.0148, ..., -0.0079, -0.0046, -0.0078], + [-0.0034, 0.0047, -0.0098, ..., -0.0081, 0.0216, -0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_2.2.after_proj.bias': tensor([-0.0038, -0.0012, 0.0049, ..., -0.0038, 0.0101, -0.0013], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_proj.weight': tensor([[-0.0121, 0.0339, 0.0119, ..., -0.0087, 0.0035, -0.0254], + [-0.0052, 0.0046, 0.0090, ..., 0.0026, 0.0165, -0.0212], + [ 0.0130, 0.0161, 0.0101, ..., -0.0052, -0.0282, 0.0151], + ..., + [-0.0036, -0.0046, 0.0041, ..., 0.0019, 0.0104, 0.0014], + [-0.0120, -0.0302, 0.0251, ..., -0.0388, 0.0052, -0.0098], + [-0.0039, -0.0037, 0.0004, ..., 0.0075, 0.0177, 0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_norm.weight': tensor([ 9.3750e-01, 1.7422e+00, 1.4844e+00, 1.5547e+00, 1.5430e-01, + 1.2207e-01, 1.7212e-02, -1.6065e-08, -2.3954e-33, 3.5553e-03, + 1.1353e-02, 1.0791e-01, -5.8594e-03, -4.5166e-03, 1.6022e-03, + -1.0490e-04, 6.5430e-02, 1.0391e+00, 1.1094e+00, 7.7820e-04, + 4.1485e-05, 1.2891e+00, 2.0312e+00, 1.8516e+00, 1.0625e+00, + 8.8281e-01, 1.3750e+00, -1.2817e-02, -6.9046e-04, 3.4714e-04, + -2.3041e-03, -2.5940e-04, 3.7354e-02, 1.1641e+00, 6.1035e-03, + 1.1172e+00, 1.6113e-02, 3.1710e-05, 1.4221e-02, -7.5378e-03, + 2.0862e-05, 1.6212e-05, 1.5020e-05, 2.0938e+00, 1.3516e+00, + 1.4922e+00, 1.5234e+00, 1.4688e+00, 9.4238e-02, 5.4932e-04, + 3.1982e-02, 2.7148e-01, 1.1719e+00, 1.1475e-01, 2.1362e-03, + 8.2520e-02, 2.8839e-03, 1.0107e-01, -4.4107e-05, 3.6163e-03, + -1.1978e-03, 5.2185e-03, -5.2490e-03, 1.1719e+00, 1.9375e+00, + 6.9531e-01, 8.5547e-01, 1.3828e+00, 4.5410e-02, -2.3193e-02, + 1.1475e-02, 3.3993e-08, -3.8037e-33, 9.4604e-03, 1.8921e-03, + -3.1586e-03, -1.5450e-04, 2.0386e-02, -4.1748e-02, 2.5635e-03, + 4.6387e-02, 4.7607e-03, 1.8188e-02, 8.7738e-04, 9.6321e-05, + -5.2185e-03, 8.4375e-01, 8.2031e-01, 1.5312e+00, 1.4609e+00, + 7.6562e-01, -5.9814e-03, -2.0885e-04, -5.9605e-05, -2.3651e-03, + 2.6894e-04, 2.1973e-02, 1.3770e-01, 2.3804e-03, 4.5654e-02, + -2.1362e-02, -8.4639e-06, 1.2891e-01, 6.4850e-05, 5.7983e-03, + -1.5736e-05, -2.9206e-05, 5.1953e-01, 1.7266e+00, 1.4922e+00, + 8.3984e-01, 7.5781e-01, 9.4727e-02, 4.0054e-04, 2.5391e-02, + 1.1562e+00, 8.0078e-02, 1.1875e+00, -7.3910e-05, 9.0332e-03, + 1.9043e-02, 4.7852e-02, 2.4438e-05, -2.1210e-03, 2.1648e-04, + -2.1729e-02, 1.1328e+00, -9.9487e-03], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.self_attn.k_proj.weight': tensor([[-1.5747e-02, 6.3171e-03, 2.8381e-03, ..., -1.1230e-02, + -3.0212e-03, 3.7842e-03], + [-1.0864e-02, -2.5635e-02, 1.5503e-02, ..., -7.3242e-03, + -4.9438e-03, -5.3406e-03], + [ 5.0964e-03, 4.3945e-03, 9.0942e-03, ..., 1.7090e-02, + -6.8665e-03, 3.7384e-03], + ..., + [-4.6082e-03, 4.9133e-03, 4.7913e-03, ..., 1.9836e-03, + -3.8605e-03, 3.3417e-03], + [-1.3550e-02, -1.5736e-05, 2.5757e-02, ..., -1.5381e-02, + 7.1716e-03, 5.3406e-03], + [ 1.9226e-03, 4.5166e-03, -7.6904e-03, ..., -4.6692e-03, + 6.8665e-03, 2.1362e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.k_norm.weight': tensor([ 8.7891e-01, 1.7109e+00, 1.5000e+00, 1.4922e+00, 1.5625e-01, + 1.2793e-01, 1.1902e-02, -6.7521e-09, -1.7935e-33, -1.6251e-03, + 1.1169e-02, 1.0889e-01, -1.0925e-02, 3.1006e-02, -2.0142e-03, + 9.8705e-05, 1.3123e-02, 1.0391e+00, 1.1094e+00, 7.6675e-04, + 7.2479e-05, 1.2891e+00, 1.7969e+00, 1.7188e+00, 1.2891e-01, + 3.8281e-01, 1.4219e+00, -1.5015e-02, 3.2425e-04, -3.0136e-04, + -2.4567e-03, 2.9564e-04, -3.6133e-02, 1.1641e+00, 2.5787e-03, + 1.1094e+00, 1.6113e-02, -3.0518e-05, -1.4954e-02, -7.6599e-03, + -1.0788e-05, 2.3365e-05, -5.0306e-05, 1.6016e+00, 6.2109e-01, + 1.4531e+00, 1.5781e+00, 1.4922e+00, -5.6885e-02, 5.0735e-04, + 1.9653e-02, -9.3994e-03, 1.1719e+00, 1.2354e-01, 2.1515e-03, + 8.2520e-02, -3.9368e-03, 1.0156e-01, -3.3855e-05, 7.0801e-02, + -1.2054e-03, 2.3315e-02, -1.5869e-02, 1.1719e+00, 1.9609e+00, + 9.9219e-01, 7.5781e-01, 1.3672e+00, 2.6855e-02, -1.0193e-02, + -2.4170e-02, 9.8255e-08, 4.4778e-33, 1.0010e-02, 1.4587e-02, + 4.3945e-02, 1.6403e-04, 3.7842e-02, -5.2490e-02, -2.6093e-03, + -4.6692e-03, 1.9287e-02, 1.5198e-02, 8.6975e-04, 9.9182e-05, + 3.2043e-03, 1.3828e+00, 1.3281e+00, 1.6250e+00, 1.5156e+00, + 3.1445e-01, 2.5177e-03, -6.5231e-04, -1.4687e-04, 2.4109e-03, + 3.3951e-04, 4.9744e-03, -2.4170e-02, -1.9684e-03, 7.1411e-03, + 2.1484e-02, -3.4332e-05, 1.2793e-01, -2.4080e-05, 5.7678e-03, + 1.9908e-05, -2.6584e-05, 1.4219e+00, 1.9531e+00, 1.4609e+00, + 1.3086e-01, 7.0703e-01, 5.0659e-03, 4.2915e-04, 4.6082e-03, + 1.1797e+00, -4.9438e-03, 1.1875e+00, 5.9366e-05, -1.8555e-02, + 1.9409e-02, 4.6387e-02, 3.0398e-05, 4.4189e-02, 2.1935e-04, + 2.8198e-02, 1.1328e+00, -1.0071e-02], dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.self_attn.v_proj.weight': tensor([[-0.0090, -0.0089, 0.0096, ..., 0.0508, -0.0099, -0.0027], + [-0.0032, 0.0112, -0.0091, ..., 0.0052, -0.0070, -0.0080], + [-0.0040, 0.0182, -0.0139, ..., -0.0025, 0.0050, 0.0128], + ..., + [ 0.0008, 0.0008, -0.0097, ..., -0.0038, 0.0085, 0.0115], + [ 0.0040, 0.0105, -0.0109, ..., 0.0032, -0.0037, -0.0028], + [-0.0021, -0.0089, 0.0081, ..., 0.0151, 0.0078, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.self_attn.output_proj.weight': tensor([[ 0.0198, 0.0126, 0.0085, ..., 0.0065, 0.0132, -0.0049], + [ 0.0198, -0.0013, 0.0054, ..., -0.0305, -0.0049, 0.0111], + [-0.0098, 0.0184, -0.0212, ..., -0.0133, -0.0063, 0.0136], + ..., + [-0.0292, -0.0199, 0.0059, ..., -0.0051, -0.0112, 0.0065], + [-0.0072, 0.0178, -0.0261, ..., -0.0131, 0.0439, 0.0035], + [-0.0305, -0.0042, 0.0276, ..., 0.0226, -0.0085, -0.0272]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_proj.weight': tensor([[-0.0057, -0.0057, 0.0170, ..., -0.0219, -0.0225, 0.0112], + [-0.0082, 0.0065, 0.0289, ..., -0.0229, -0.0029, 0.0105], + [ 0.0126, -0.0037, -0.0076, ..., 0.0052, 0.0110, -0.0043], + ..., + [ 0.0104, 0.0056, -0.0078, ..., -0.0189, -0.0057, 0.0044], + [ 0.0064, 0.0035, -0.0112, ..., 0.0045, -0.0034, 0.0016], + [-0.0064, -0.0037, 0.0124, ..., 0.0123, -0.0010, 0.0044]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_norm.weight': tensor([0.0967, 0.0156, 0.4414, 0.0576, 0.1416, 0.0320, 0.1377, 0.1201, 0.5195, + 0.0297, 0.0488, 0.0168, 0.0378, 0.0713, 0.0508, 0.1836, 0.4746, 0.0219, + 0.7578, 0.0405, 0.0283, 0.1533, 0.0262, 0.0757, 0.0425, 0.3906, 0.5156, + 0.2617, 0.0684, 0.0405, 0.0231, 0.2178, 0.6797, 0.2402, 0.0322, 0.0349, + 0.0141, 0.0928, 0.0175, 0.0359, 0.0491, 0.0986, 0.0206, 0.0298, 0.2041, + 0.0337, 0.3828, 0.0153, 0.0413, 0.0170, 0.0173, 0.4180, 0.4668, 0.3418, + 0.0708, 0.6211, 0.5977, 0.0447, 0.1060, 0.0198, 0.0767, 0.4648, 0.0071, + 0.5000, 0.1475, 0.3926, 0.0815, 0.0471, 0.0630, 0.4473, 0.0280, 0.4707, + 0.0135, 0.0366, 0.1377, 0.0698, 0.0148, 0.0991, 0.6523, 0.0337, 0.0732, + 0.0908, 0.0388, 0.0503, 0.4492, 0.0601, 0.0640, 0.0879, 0.0189, 0.1396, + 0.0327, 0.1758, 0.0408, 0.0549, 0.0376, 0.0996, 0.0088, 0.3340, 0.5195, + 0.0344, 0.0520, 0.2949, 0.0322, 0.1108, 0.0325, 0.2158, 0.0123, 0.6406, + 0.6211, 0.0175, 0.0161, 0.7188, 0.0410, 0.0160, 0.4805, 0.2090, 0.2695, + 0.6992, 0.0332, 0.0762, 0.0664, 0.5586, 0.6133, 0.0234, 0.0225, 0.1855, + 0.0366, 0.0369], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_proj.weight': tensor([[-3.1586e-03, 1.8158e-03, 4.1504e-03, ..., -1.4114e-03, + 1.0681e-02, 3.1281e-03], + [ 2.8076e-03, -6.7711e-05, -1.6861e-03, ..., 4.7913e-03, + -1.1108e-02, 5.2795e-03], + [-1.9043e-02, 7.6904e-03, -3.2043e-04, ..., 1.8433e-02, + 1.0681e-02, 3.5553e-03], + ..., + [ 5.8594e-03, 3.1128e-03, -5.7983e-03, ..., 9.4986e-04, + -1.9287e-02, 5.4932e-04], + [-9.5825e-03, -1.8539e-03, -1.0681e-02, ..., 6.0425e-03, + -1.0071e-02, 5.9509e-03], + [ 2.0264e-02, -1.4038e-03, 1.0681e-02, ..., -8.6670e-03, + 7.0190e-03, -6.2866e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_norm.weight': tensor([0.0967, 0.0157, 0.4414, 0.0574, 0.1416, 0.0327, 0.1387, 0.1206, 0.5195, + 0.0302, 0.0498, 0.0168, 0.0378, 0.0713, 0.0508, 0.1855, 0.4746, 0.0217, + 0.7539, 0.0403, 0.0281, 0.1533, 0.0264, 0.0752, 0.0422, 0.3867, 0.5156, + 0.2598, 0.0688, 0.0408, 0.0238, 0.2197, 0.6797, 0.2402, 0.0330, 0.0349, + 0.0137, 0.0928, 0.0175, 0.0354, 0.0483, 0.0996, 0.0210, 0.0303, 0.2051, + 0.0337, 0.3828, 0.0154, 0.0410, 0.0166, 0.0179, 0.4180, 0.4668, 0.3418, + 0.0708, 0.6211, 0.6016, 0.0439, 0.1050, 0.0201, 0.0767, 0.4648, 0.0075, + 0.5000, 0.1465, 0.3926, 0.0815, 0.0471, 0.0630, 0.4473, 0.0282, 0.4707, + 0.0132, 0.0366, 0.1377, 0.0693, 0.0148, 0.0996, 0.6523, 0.0334, 0.0723, + 0.0908, 0.0391, 0.0503, 0.4492, 0.0603, 0.0630, 0.0874, 0.0194, 0.1416, + 0.0337, 0.1748, 0.0408, 0.0552, 0.0378, 0.1006, 0.0087, 0.3340, 0.5195, + 0.0337, 0.0525, 0.2949, 0.0325, 0.1108, 0.0325, 0.2158, 0.0122, 0.6406, + 0.6211, 0.0172, 0.0162, 0.7188, 0.0408, 0.0161, 0.4824, 0.2090, 0.2695, + 0.6953, 0.0327, 0.0747, 0.0664, 0.5586, 0.6133, 0.0233, 0.0225, 0.1855, + 0.0378, 0.0374], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.v_proj.weight': tensor([[ 0.0226, 0.0038, -0.0243, ..., 0.0069, -0.0026, -0.0189], + [-0.0203, 0.0140, 0.0005, ..., 0.0066, -0.0025, -0.0311], + [ 0.0251, -0.0085, -0.0001, ..., -0.0010, -0.0099, -0.0452], + ..., + [ 0.0023, -0.0017, 0.0232, ..., -0.0073, -0.0212, 0.0034], + [-0.0327, -0.0032, -0.0256, ..., -0.0060, 0.0122, -0.0121], + [-0.0164, 0.0014, 0.0045, ..., -0.0046, -0.0042, 0.0089]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.output_proj.weight': tensor([[ 0.0017, 0.0038, -0.0017, ..., -0.0120, 0.0074, 0.0096], + [-0.0315, -0.0042, -0.0132, ..., -0.0078, -0.0199, -0.0077], + [ 0.0161, -0.0069, 0.0054, ..., -0.0117, 0.0118, -0.0172], + ..., + [ 0.0181, -0.0146, 0.0189, ..., 0.0119, -0.0060, -0.0081], + [-0.0137, -0.0216, 0.0060, ..., 0.0129, -0.0192, 0.0261], + [ 0.0005, -0.0215, 0.0315, ..., -0.0364, 0.0332, 0.0188]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0057, -0.0065, ..., 0.0266, -0.0278, 0.0075], + ..., + [-0.0217, -0.0410, 0.0488, ..., -0.0039, -0.0003, -0.0255], + [ 0.0036, 0.0066, 0.0013, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0059, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0101, 0.0038, ..., -0.0236, -0.0109, -0.0248], + [-0.0079, -0.0013, 0.0293, ..., 0.0072, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0143, 0.0126, -0.0094], + ..., + [-0.0035, 0.0320, -0.0233, ..., -0.0060, -0.0015, 0.0188], + [ 0.0136, -0.0352, 0.0267, ..., -0.0305, -0.0045, 0.0037], + [-0.0205, 0.0140, 0.0025, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img.weight': tensor([[-0.0079, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0029, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0057, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0018], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, -0.0027, -0.0020], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0023, ..., -0.0103, 0.0074, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_2.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_2.3.mlp.layer1.weight': tensor([[ 0.0151, -0.0038, 0.0084, ..., 0.0068, 0.0133, -0.0104], + [-0.0098, 0.0107, 0.0242, ..., 0.0029, 0.0115, -0.0012], + [-0.0037, 0.0354, -0.0305, ..., 0.0278, 0.0125, 0.0047], + ..., + [-0.0060, 0.0233, 0.0693, ..., -0.0088, 0.0012, -0.0090], + [-0.0055, -0.0082, -0.0050, ..., -0.0095, -0.0016, -0.0164], + [-0.0005, 0.0047, -0.0097, ..., -0.0092, -0.0284, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.mlp.layer2.weight': tensor([[ 0.0067, 0.0117, 0.0032, ..., -0.0220, -0.0221, -0.0071], + [ 0.0104, 0.0247, 0.0068, ..., -0.0092, 0.0356, -0.0181], + [ 0.0040, -0.0050, -0.0037, ..., 0.0150, 0.0275, 0.0042], + ..., + [ 0.0045, 0.0088, 0.0064, ..., -0.0083, -0.0062, 0.0048], + [ 0.0051, -0.0092, -0.0031, ..., 0.0077, 0.0183, 0.0037], + [-0.0036, 0.0048, 0.0170, ..., 0.0052, 0.0349, 0.0073]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0226, -0.0222, -0.0220, ..., 0.0010, 0.0008, 0.0009], + [-0.0315, -0.0306, -0.0298, ..., 0.0026, 0.0027, 0.0027], + [-0.0222, -0.0214, -0.0205, ..., 0.0021, 0.0023, 0.0023], + ..., + [ 0.0605, 0.0588, 0.0569, ..., -0.0037, -0.0037, -0.0037], + [ 0.0496, 0.0479, 0.0464, ..., -0.0032, -0.0031, -0.0030], + [-0.0256, -0.0250, -0.0248, ..., 0.0022, 0.0023, 0.0024]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0061, 0.0044, 0.0062, ..., 0.0099, 0.0130, 0.0059], + [ 0.0005, 0.0014, 0.0015, ..., -0.0015, 0.0004, 0.0016], + [ 0.0015, -0.0013, 0.0005, ..., 0.0037, 0.0038, -0.0014], + ..., + [-0.0237, -0.0762, -0.0615, ..., 0.0369, 0.0354, -0.0693], + [ 0.0089, 0.0082, 0.0085, ..., 0.0182, 0.0150, 0.0092], + [ 0.0164, 0.0099, 0.0124, ..., 0.0187, 0.0208, 0.0113]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_cross_attn.1.weight': tensor([[-8.2031e-02, -8.0078e-02, -7.7637e-02, ..., -1.0443e-04, + -1.2016e-04, -1.0967e-04], + [ 3.0029e-02, 2.9419e-02, 2.8564e-02, ..., 5.6624e-06, + 3.1948e-05, 1.5163e-04], + [ 3.2715e-02, 3.1494e-02, 3.0518e-02, ..., 2.6703e-04, + 2.8992e-04, 1.9836e-04], + ..., + [ 5.5908e-02, 5.4199e-02, 5.3223e-02, ..., -1.8978e-04, + -1.8501e-04, -1.1683e-04], + [ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.8311e-04, + -1.2112e-04, -1.7071e-04], + [-3.0884e-02, -2.9907e-02, -2.9175e-02, ..., -5.8889e-05, + 9.3460e-05, 1.1492e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_cross_attn.2.weight': tensor([[-7.9346e-03, 3.7842e-03, -8.8501e-03, ..., 2.8320e-02, + 3.0396e-02, 3.6926e-03], + [ 6.7139e-03, -9.3994e-03, 1.1841e-02, ..., -2.4605e-04, + 4.5166e-03, 4.6387e-03], + [ 2.0905e-03, -1.2207e-03, 4.6997e-03, ..., -7.9346e-03, + -6.3477e-03, -4.7913e-03], + ..., + [-1.5991e-02, 4.6387e-03, 3.5400e-02, ..., 2.5146e-02, + 3.2227e-02, -8.7891e-03], + [-2.3315e-02, 1.2085e-02, 3.6865e-02, ..., 1.0254e-02, + 1.5869e-02, -1.4099e-02], + [ 1.3977e-02, -8.5449e-03, -7.0801e-03, ..., 1.5717e-03, + 1.6098e-03, -4.5300e-05]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_mlp.1.weight': tensor([[-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., 1.3638e-04, + 2.5940e-04, 2.0313e-04], + [ 1.3245e-02, 1.1841e-02, 1.1597e-02, ..., 3.7193e-04, + 4.5204e-04, 4.4441e-04], + [-7.3242e-02, -7.1777e-02, -7.0801e-02, ..., -4.4823e-04, + -3.7766e-04, -4.1199e-04], + ..., + [-6.5430e-02, -6.4453e-02, -6.3477e-02, ..., 1.6785e-04, + 2.7084e-04, 1.6785e-04], + [-4.4922e-02, -4.4434e-02, -4.4189e-02, ..., 2.0981e-04, + 6.8188e-05, 7.9632e-05], + [ 6.2012e-02, 6.1035e-02, 6.0791e-02, ..., -1.4210e-04, + -1.6975e-04, -1.3351e-04]], dtype=torch.bfloat16), 'net.control_blocks_2.3.adaln_modulation_mlp.2.weight': tensor([[-0.0141, 0.0030, -0.0100, ..., -0.0164, -0.0199, 0.0178], + [-0.0017, -0.0063, 0.0064, ..., -0.0015, 0.0049, -0.0028], + [-0.0016, -0.0029, -0.0036, ..., -0.0010, -0.0017, 0.0016], + ..., + [-0.0493, -0.0312, -0.0461, ..., -0.0361, -0.0109, 0.0266], + [ 0.0075, -0.0177, 0.0260, ..., 0.0075, 0.0052, -0.0075], + [ 0.0135, 0.0118, 0.0070, ..., 0.0121, 0.0072, -0.0096]], + dtype=torch.bfloat16), 'net.control_blocks_2.3.after_proj.weight': tensor([[ 2.8229e-03, 1.6724e-02, -1.8311e-02, ..., 7.8125e-03, + 3.9368e-03, -6.8970e-03], + [-8.4229e-03, -3.6316e-03, 7.6599e-03, ..., -9.5844e-05, + 1.6235e-02, -8.6060e-03], + [ 9.6436e-03, 4.3335e-03, -3.9978e-03, ..., 3.8452e-03, + 2.1820e-03, -7.6599e-03], + ..., + [ 1.7456e-02, 1.2329e-02, -5.0354e-03, ..., 7.7515e-03, + -6.1035e-03, 5.2795e-03], + [-6.2866e-03, 7.4463e-03, -3.2349e-03, ..., 1.0437e-02, + -3.9673e-03, 4.2419e-03], + [ 3.4332e-03, -1.4832e-02, -5.6458e-03, ..., 2.1362e-02, + -1.2268e-02, 8.9722e-03]], dtype=torch.bfloat16), 'net.control_blocks_2.3.after_proj.bias': tensor([ 0.0085, -0.0193, -0.0089, ..., -0.0041, -0.0183, 0.0126], + dtype=torch.bfloat16)} +[06-26 07:20:46|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_2.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:46|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.3.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_3.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_3.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 07:20:46|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 2 checkpoint. +[06-26 07:20:46|DEBUG|cosmos_transfer2/_src/imaginaire/utils/checkpoint_db.py:344:download] Resolving checkpoint nvidia/Cosmos-Transfer2.5-2B/general/seg(5136ef49-6d8d-42e8-8abf-7dac722a304a) +[06-26 07:20:46|CRITICAL|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:694:load_multi_branch_checkpoints] Start loading checkpoint for control branch 3 from /scratch/ssaha/.cache/huggingface/hub/models--nvidia--Cosmos-Transfer2.5-2B/snapshots/23057a4167b89de89a4a397fdbf3887994d115eb/general/seg/5136ef49-6d8d-42e8-8abf-7dac722a304a_ema_bf16.pt +[06-26 07:20:52|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:733:load_multi_branch_checkpoints] Checkpoint to model keys: {'net.accum_video_sample_counter': tensor(4575657222473777152), 'net.accum_image_sample_counter': tensor(4575657222473777152), 'net.accum_iteration': tensor(4575657222473777152), 'net.accum_train_in_hours': tensor(1., dtype=torch.bfloat16), 'net.x_embedder.proj.1.weight': tensor([[-0.0038, -0.0085, -0.0181, ..., 0.0075, -0.0039, -0.0038], + [ 0.0035, 0.0008, 0.0005, ..., -0.0284, 0.0127, -0.0237], + [ 0.0221, 0.0211, 0.0220, ..., -0.0183, 0.0103, 0.0027], + ..., + [ 0.0415, 0.0417, 0.0466, ..., 0.0079, -0.0136, 0.0092], + [ 0.0002, 0.0030, -0.0051, ..., -0.0566, -0.0515, -0.0825], + [-0.0140, -0.0126, -0.0109, ..., 0.0115, -0.0095, -0.0031]], + dtype=torch.bfloat16), 'net.pos_embedder.seq': tensor([ 0., 1., 2., 3., 4., 5., 6., 7., 8., 9., 10., 11., + 12., 13., 14., 15., 16., 17., 18., 19., 20., 21., 22., 23., + 24., 25., 26., 27., 28., 29., 30., 31., 32., 33., 34., 35., + 36., 37., 38., 39., 40., 41., 42., 43., 44., 45., 46., 47., + 48., 49., 50., 51., 52., 53., 54., 55., 56., 57., 58., 59., + 60., 61., 62., 63., 64., 65., 66., 67., 68., 69., 70., 71., + 72., 73., 74., 75., 76., 77., 78., 79., 80., 81., 82., 83., + 84., 85., 86., 87., 88., 89., 90., 91., 92., 93., 94., 95., + 96., 97., 98., 99., 100., 101., 102., 103., 104., 105., 106., 107., + 108., 109., 110., 111., 112., 113., 114., 115., 116., 117., 118., 119., + 120., 121., 122., 123., 124., 125., 126., 127.], dtype=torch.bfloat16), 'net.pos_embedder.dim_spatial_range': tensor([0.0000, 0.0476, 0.0952, 0.1426, 0.1904, 0.2383, 0.2852, 0.3340, 0.3809, + 0.4277, 0.4766, 0.5234, 0.5703, 0.6172, 0.6680, 0.7148, 0.7617, 0.8086, + 0.8555, 0.9062, 0.9531], dtype=torch.bfloat16), 'net.pos_embedder.dim_temporal_range': tensor([0.0000, 0.0454, 0.0908, 0.1367, 0.1816, 0.2275, 0.2734, 0.3184, 0.3633, + 0.4082, 0.4551, 0.5000, 0.5469, 0.5898, 0.6367, 0.6836, 0.7266, 0.7734, + 0.8164, 0.8633, 0.9102, 0.9531], dtype=torch.bfloat16), 'net.t_embedder.1.linear_1.weight': tensor([[-1.7929e-03, -1.7853e-03, -1.7700e-03, ..., -1.5564e-03, + -1.3809e-03, -1.5030e-03], + [-1.6113e-02, -1.6113e-02, -1.6113e-02, ..., 5.0068e-05, + 9.5963e-06, 1.1206e-04], + [-1.9989e-03, -6.4850e-04, -5.7983e-04, ..., -1.3199e-03, + -1.3123e-03, -1.2283e-03], + ..., + [-2.1973e-02, -2.1973e-02, -2.1606e-02, ..., -2.8419e-04, + -2.8229e-04, -1.6212e-04], + [-4.5776e-04, -5.0354e-04, -6.7139e-04, ..., -1.7090e-03, + -1.6174e-03, -1.6251e-03], + [ 2.1935e-04, 1.6308e-04, 2.9755e-04, ..., -1.2589e-03, + -1.1902e-03, -1.3199e-03]], dtype=torch.bfloat16), 'net.t_embedder.1.linear_2.weight': tensor([[-1.9379e-03, 7.7057e-04, -1.6861e-03, ..., 9.9945e-04, + -4.4632e-04, -3.1586e-03], + [-2.9602e-03, -1.5926e-04, 1.6556e-03, ..., -7.0572e-05, + -3.0060e-03, 4.2152e-04], + [-1.4725e-03, -8.2016e-05, -4.6921e-04, ..., -3.8624e-05, + 7.0190e-04, -2.3041e-03], + ..., + [ 1.8921e-03, -1.6708e-03, 6.0201e-06, ..., -3.4790e-03, + -9.3079e-04, 3.8147e-04], + [ 2.5482e-03, 2.6855e-03, 6.6528e-03, ..., 3.0060e-03, + -2.3651e-03, -8.1062e-05], + [-1.2970e-03, -4.0283e-03, -1.4496e-03, ..., -5.4932e-03, + -1.2436e-03, -4.0283e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_proj.weight': tensor([[-0.0016, 0.0070, 0.0106, ..., -0.0017, 0.0031, 0.0056], + [-0.0002, 0.0019, -0.0023, ..., -0.0072, -0.0018, -0.0023], + [-0.0005, -0.0008, 0.0010, ..., -0.0039, -0.0056, -0.0038], + ..., + [-0.0085, -0.0052, -0.0018, ..., 0.0072, 0.0019, -0.0107], + [-0.0006, -0.0092, -0.0101, ..., -0.0092, -0.0039, -0.0054], + [ 0.0039, -0.0067, 0.0009, ..., 0.0075, -0.0037, -0.0064]], + dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm.weight': tensor([ 9.3750e-02, -3.8757e-03, 1.5991e-02, 9.8145e-02, -5.9903e-06, + 3.3379e-05, -2.0146e-05, 2.6226e-06, -2.1219e-05, -1.4901e-05, + -1.0777e-04, 2.0981e-05, -5.8413e-06, 3.6322e-07, -1.1828e-07, + 8.5682e-07, 1.2666e-06, -4.5300e-06, -1.3225e-07, -4.5300e-06, + 1.4454e-06, -5.5790e-05, 1.9165e-02, -6.0350e-07, -1.4648e-03, + 2.8906e-01, -4.9591e-04, -2.3079e-04, 3.1982e-02, 7.1289e-02, + 9.2015e-07, 3.2902e-05, -2.1076e-04, -1.9222e-06, 3.3677e-06, + 1.7643e-05, 6.6683e-07, 5.5075e-05, 2.7474e-08, 2.5481e-06, + 6.7230e-09, 3.5070e-09, -1.7524e-05, 6.5918e-03, 3.1982e-02, + -2.7865e-06, 1.6174e-03, 1.7929e-03, 1.0986e-01, -2.8133e-05, + -5.2214e-05, -3.3379e-06, 5.4389e-07, 1.4380e-06, -1.2636e-05, + -1.8105e-06, 2.1607e-06, 1.9968e-06, -2.1011e-06, -2.0117e-06, + -3.0845e-06, 8.7544e-08, -6.5029e-11, -1.8772e-09, 9.7168e-02, + 9.4238e-02, 2.6758e-01, 1.1749e-03, -9.6112e-07, 1.2815e-05, + -2.1696e-05, 8.1062e-06, -8.7023e-06, -1.2493e-04, -2.4676e-05, + -1.0431e-06, -6.6310e-07, -3.7067e-07, 8.1956e-07, 6.4448e-07, + 5.4121e-05, 5.8860e-07, 8.4750e-08, 9.4175e-06, 6.1512e-05, + -1.4603e-06, -2.2888e-03, -1.6168e-06, 2.9297e-01, 2.0294e-03, + 2.2266e-01, 7.3242e-03, 1.7624e-03, -5.4121e-05, -9.5367e-06, + -1.5050e-06, -1.3232e-05, -4.2142e-08, 4.1962e-05, 6.9141e-06, + -1.5199e-05, 5.1223e-09, 1.4544e-05, -2.3007e-05, 1.3188e-06, + 4.5896e-06, -1.3737e-08, 2.2705e-02, -5.4550e-04, 2.5332e-06, + 2.8711e-01, 1.5918e-01, 4.4441e-04, 6.8665e-05, 6.9141e-06, + 2.0981e-05, -9.1195e-06, -1.1921e-04, -3.1662e-04, -4.5061e-05, + 1.0014e-04, -1.1623e-06, -5.1558e-06, 2.8126e-07, -9.4771e-06, + -1.3113e-05, 1.5018e-08, -8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.k_proj.weight': tensor([[-6.5918e-03, 7.1716e-03, -7.1411e-03, ..., 2.7313e-03, + -1.3173e-05, 3.2196e-03], + [-1.5503e-02, -2.0885e-04, 5.5847e-03, ..., 1.8616e-03, + -9.9945e-04, 2.3804e-03], + [ 2.9449e-03, -3.8910e-03, 1.3809e-03, ..., -2.3365e-04, + -1.1969e-04, 9.5825e-03], + ..., + [ 3.2959e-03, 2.8381e-03, -4.1199e-03, ..., -4.6387e-03, + -1.2360e-03, -3.2997e-04], + [ 3.4332e-03, 2.7313e-03, -2.5482e-03, ..., -3.0518e-03, + 1.0605e-03, 1.8005e-03], + [ 3.4180e-03, -4.1771e-04, -1.5030e-03, ..., -3.2196e-03, + -2.7847e-04, -3.1281e-04]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm.weight': tensor([ 9.6680e-02, 3.7598e-02, 6.9580e-03, 9.8145e-02, -6.5267e-06, + -3.3617e-05, 1.9550e-05, -4.3586e-07, 6.3896e-05, -2.2411e-05, + -1.0729e-04, -9.7752e-06, 5.6624e-06, -3.7439e-07, 1.1548e-07, + 8.9034e-07, -1.4752e-06, -1.5736e-05, 2.1458e-06, 4.5300e-06, + 1.5274e-06, 5.6028e-05, 1.2939e-02, -1.6484e-07, 1.6403e-04, + 2.8906e-01, 6.3705e-04, -2.7618e-03, 2.2888e-04, 1.8358e-05, + 1.0245e-08, -3.1471e-05, 2.0885e-04, -1.9372e-06, -5.0962e-06, + 1.8001e-05, -3.4273e-07, -5.4359e-05, 1.6415e-08, 2.6971e-06, + 3.9290e-09, -3.6089e-09, -1.7524e-05, 1.6235e-02, 3.0640e-02, + -5.9605e-07, 3.7598e-02, 1.0303e-01, 1.3184e-01, -9.5367e-05, + -4.6253e-05, 2.3395e-06, 1.1623e-06, 1.6019e-06, 4.7088e-06, + 9.6858e-07, -4.4405e-06, 4.0531e-06, 3.2187e-06, 1.9968e-06, + 8.3447e-06, 1.2107e-07, 4.1609e-11, 3.4197e-09, 3.5477e-04, + 9.3262e-02, 2.6758e-01, -3.7354e-02, -3.0845e-06, -4.2617e-06, + 2.1458e-05, -7.1526e-06, -1.1027e-05, 1.2493e-04, -2.4319e-05, + -4.3213e-07, -5.4389e-07, -3.7439e-07, -8.3447e-07, 9.8953e-09, + -6.4850e-05, -2.3516e-08, -1.3597e-07, -9.2983e-06, 1.5736e-05, + 1.0252e-05, 1.0223e-03, -5.8115e-06, 2.8906e-01, -4.0283e-03, + 2.2363e-01, -6.7902e-04, 3.1738e-02, 7.0801e-02, 8.9407e-06, + -3.4571e-06, -3.6657e-06, 5.2387e-08, 3.3617e-05, -6.9737e-06, + -1.5497e-05, -1.4529e-07, -1.4901e-05, 2.3127e-05, 1.3188e-06, + 4.5896e-06, 8.6147e-09, 1.6724e-02, 1.0376e-02, 3.4422e-06, + 2.9102e-01, 1.3477e-01, 8.3008e-02, -1.0490e-05, 3.0160e-05, + -2.0981e-05, 3.2663e-05, 1.1921e-04, 3.3188e-04, -4.5061e-05, + -7.6771e-05, 2.4028e-07, 5.0366e-06, -2.7753e-07, -9.1195e-06, + -1.3113e-05, 1.6391e-07, 8.8215e-06], dtype=torch.bfloat16), 'net.blocks.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.self_attn.v_proj.weight': tensor([[ 4.6997e-03, 6.5613e-03, 1.9226e-03, ..., -1.6632e-03, + -2.4986e-04, -1.4038e-03], + [ 2.9755e-03, 5.7068e-03, -2.4261e-03, ..., 5.0964e-03, + 1.8768e-03, -4.6997e-03], + [ 1.3611e-02, 2.1362e-03, 2.9755e-03, ..., -2.4986e-04, + -1.8463e-03, 5.4016e-03], + ..., + [ 9.6130e-04, -2.0294e-03, 4.0894e-03, ..., 5.5695e-04, + 1.3542e-04, -1.6937e-03], + [ 1.2207e-03, 3.3112e-03, 1.0376e-03, ..., 9.0027e-04, + 1.3199e-03, -9.5749e-04], + [-4.5471e-03, 3.5858e-03, -1.8835e-05, ..., 9.4223e-04, + -2.1057e-03, -4.7913e-03]], dtype=torch.bfloat16), 'net.blocks.0.self_attn.output_proj.weight': tensor([[ 0.0178, -0.0322, -0.0254, ..., -0.0014, 0.0018, 0.0017], + [ 0.0121, -0.0075, 0.0114, ..., 0.0006, -0.0008, -0.0014], + [-0.0018, -0.0003, -0.0042, ..., -0.0022, -0.0014, 0.0031], + ..., + [ 0.0037, 0.0010, 0.0074, ..., 0.0024, -0.0015, -0.0025], + [-0.0092, 0.0026, 0.0039, ..., -0.0039, 0.0020, 0.0090], + [-0.0038, 0.0069, 0.0007, ..., 0.0040, 0.0029, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_proj.weight': tensor([[-6.7139e-04, -4.6921e-04, -2.5635e-03, ..., -6.9427e-04, + -1.7471e-03, -2.7771e-03], + [-1.5717e-03, -1.7452e-04, 3.2043e-03, ..., 6.8188e-05, + 1.8692e-03, 2.9182e-04], + [-4.4632e-04, 2.4109e-03, 2.1362e-03, ..., 1.4267e-03, + 5.2490e-03, 8.0490e-04], + ..., + [ 3.2654e-03, 2.6398e-03, 9.2316e-04, ..., -4.2152e-04, + -6.0120e-03, 6.2256e-03], + [-3.0365e-03, -1.0452e-03, -2.5635e-03, ..., -1.9836e-03, + 4.6997e-03, -5.5237e-03], + [-3.8385e-05, -3.1281e-03, 2.0905e-03, ..., -2.0447e-03, + 2.4872e-03, 2.5482e-03]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm.weight': tensor([ 1.4191e-03, 2.9907e-03, -3.2187e-05, 4.1199e-03, 4.2439e-05, + -4.0531e-05, -5.8413e-06, 6.1340e-03, 4.5166e-02, 2.6978e-02, + -6.5775e-09, 3.4912e-02, 5.4297e-01, 2.1210e-03, 2.3315e-02, + 9.7275e-04, 3.8477e-01, -5.3883e-05, 1.9379e-03, 5.8594e-03, + 6.1340e-03, 1.8082e-03, 9.6191e-02, 6.9580e-03, 6.9580e-03, + 5.1880e-03, 2.3438e-02, 4.9174e-06, 1.0498e-02, 1.8311e-03, + 3.0664e-01, 1.7578e-02, 2.5272e-05, 3.5156e-01, 5.2734e-01, + 5.4932e-03, 1.0315e-02, 3.6377e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.1117e-04, -2.1648e-04, 8.8501e-03, 1.6937e-03, + 4.7119e-02, 3.5858e-03, 1.3046e-03, 2.5177e-03, 1.5450e-04, + 1.4648e-03, -9.6321e-05, -8.5235e-06, 1.0681e-03, 2.8253e-05, + 3.2196e-03, 1.2085e-02, 6.9885e-03, 1.3367e-02, 8.0566e-03, + 2.8419e-04, 3.0756e-05, 8.9111e-03, 2.1057e-03, 3.4332e-05, + 1.7166e-03, 1.6022e-04, 2.0801e-01, 1.7700e-03, 1.4771e-02, + 3.5352e-01, -1.5497e-05, 4.3701e-02, 1.0193e-02, 1.4067e-05, + 1.8539e-03, 6.0730e-03, 2.6093e-03, 2.0294e-03, 8.5449e-03, + 6.0120e-03, 7.1106e-03, 1.5332e-01, 3.2654e-03, 1.8845e-03, + 1.7212e-02, -7.5912e-04, 1.6212e-05, 3.8147e-04, 4.6094e-01, + 8.1177e-03, 1.8539e-03, 2.0142e-02, 2.1210e-03, -1.4246e-05, + 3.5742e-01, 2.0790e-04, 3.4943e-03, -3.1739e-06, 9.1553e-04, + 2.9883e-01, -1.9550e-05, 4.9591e-04, 1.9836e-03, 1.8311e-02, + -4.7445e-05, 5.4321e-03, -5.2154e-07, 5.2185e-03, 7.2021e-03, + 6.2866e-03, 1.1749e-03, 1.1475e-02, 5.8594e-03, 1.2390e-02, + 6.6528e-03, 1.5015e-02, 1.6479e-03, 1.8768e-03, 4.3945e-03, + 8.9169e-05, 6.0730e-03, 8.6060e-03, 5.9891e-04, 3.8910e-03, + 7.7438e-04, 1.9531e-03, 1.1597e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_proj.weight': tensor([[-0.0002, 0.0006, 0.0024, ..., -0.0010, -0.0016, -0.0012], + [-0.0049, -0.0019, -0.0047, ..., -0.0025, -0.0025, -0.0026], + [-0.0019, -0.0013, -0.0029, ..., -0.0027, -0.0051, -0.0022], + ..., + [ 0.0038, -0.0023, 0.0025, ..., 0.0014, 0.0016, -0.0009], + [ 0.0019, 0.0004, -0.0025, ..., -0.0034, 0.0003, 0.0019], + [ 0.0015, 0.0020, -0.0031, ..., -0.0040, 0.0008, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm.weight': tensor([ 1.3962e-03, 3.0060e-03, -3.1233e-05, 3.9062e-03, 5.3167e-05, + 1.1730e-04, 9.1195e-06, 6.1035e-03, 4.4189e-02, 2.7100e-02, + -3.2783e-07, 3.4912e-02, 5.4688e-01, 2.0752e-03, 2.3193e-02, + 9.4986e-04, 3.8477e-01, -5.3644e-05, 1.8616e-03, 5.7068e-03, + 6.2256e-03, 1.8616e-03, 9.6191e-02, 7.0190e-03, 6.8054e-03, + 5.1575e-03, 2.3315e-02, 4.8578e-06, 9.7046e-03, 1.8234e-03, + 3.0664e-01, 1.7334e-02, 2.5392e-05, 3.5156e-01, 5.2344e-01, + 5.4016e-03, 1.0437e-02, 3.6621e-02, 1.3184e-01, 1.3733e-02, + 3.1494e-02, 5.3787e-04, -2.1935e-04, 8.7891e-03, 9.3460e-04, + 4.6875e-02, 3.5858e-03, 1.2741e-03, 2.5177e-03, 1.5450e-04, + 1.4191e-03, -1.0061e-04, -7.5996e-06, 1.0681e-03, 9.4986e-04, + 3.2043e-03, 1.1902e-02, 6.9580e-03, 1.3672e-02, 8.0566e-03, + 1.8311e-04, 3.0518e-05, 8.7280e-03, 2.0752e-03, 3.0518e-05, + 1.7548e-03, 1.7548e-04, 2.0801e-01, 1.7548e-03, 1.4771e-02, + 3.5352e-01, -1.4901e-05, 4.2236e-02, 1.0254e-02, 1.3888e-05, + 2.1210e-03, 5.2795e-03, 2.6245e-03, 2.0142e-03, 8.4839e-03, + 6.0425e-03, 7.1716e-03, 1.5332e-01, 3.0365e-03, 1.8845e-03, + 1.7212e-02, 2.1362e-03, 1.5378e-05, 2.9373e-04, 4.6094e-01, + 8.1177e-03, 1.7929e-03, 1.9897e-02, 2.1210e-03, -5.2452e-06, + 3.5742e-01, -1.7834e-04, 3.4943e-03, 8.9779e-07, 8.8882e-04, + 2.9883e-01, -2.8908e-06, 2.0905e-03, 1.9684e-03, 1.8433e-02, + -4.7445e-05, 5.4321e-03, -4.9174e-07, 4.5776e-03, 7.1411e-03, + 6.4392e-03, 1.1902e-03, 1.1475e-02, 5.7983e-03, 1.2634e-02, + 6.5613e-03, 1.4954e-02, 1.7395e-03, 2.5330e-03, 4.4556e-03, + 2.9325e-05, 6.1951e-03, 8.6060e-03, 4.6921e-04, 3.1891e-03, + 6.9046e-04, 1.9836e-03, 1.0223e-03], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.v_proj.weight': tensor([[-1.0071e-02, 1.4954e-03, 1.1047e-02, ..., 2.3804e-03, + 2.8534e-03, -1.1169e-02], + [-2.2095e-02, 5.4550e-04, 7.4768e-03, ..., -3.3417e-03, + -8.1787e-03, -7.5912e-04], + [-1.0986e-02, 3.0060e-03, 5.9814e-03, ..., -1.4591e-04, + 1.1902e-02, -1.1536e-02], + ..., + [ 8.7891e-03, 6.8283e-04, 2.8687e-03, ..., 3.7432e-05, + 5.9204e-03, 1.9409e-02], + [-6.9885e-03, 2.6245e-03, 1.1353e-02, ..., 1.8234e-03, + -1.7944e-02, -2.3804e-03], + [-1.7334e-02, -3.4332e-03, 6.3782e-03, ..., -3.8300e-03, + -8.4839e-03, -2.2339e-02]], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.output_proj.weight': tensor([[ 0.0070, -0.0005, 0.0078, ..., -0.0048, 0.0012, 0.0107], + [ 0.0007, 0.0075, -0.0077, ..., 0.0111, -0.0046, -0.0034], + [ 0.0011, 0.0009, -0.0016, ..., 0.0055, -0.0015, -0.0033], + ..., + [-0.0032, -0.0014, -0.0007, ..., 0.0011, -0.0022, -0.0009], + [ 0.0140, -0.0027, 0.0084, ..., -0.0080, -0.0115, -0.0018], + [ 0.0128, -0.0123, -0.0051, ..., -0.0069, -0.0013, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img.weight': tensor([[ 0.0098, 0.0635, 0.0159, ..., 0.0143, 0.0022, 0.0062], + [-0.0022, 0.0054, -0.0327, ..., -0.0080, 0.0010, 0.0298], + [-0.0234, 0.0142, -0.0282, ..., 0.0134, -0.0026, 0.0283], + ..., + [-0.0026, 0.0125, 0.0248, ..., 0.0437, -0.0073, -0.0192], + [ 0.0277, 0.0148, 0.0054, ..., -0.0015, -0.0091, 0.0146], + [ 0.0437, 0.0366, 0.0082, ..., 0.0016, -0.0129, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.v_img.weight': tensor([[-0.0006, -0.0310, 0.0300, ..., -0.0067, -0.0043, -0.0339], + [-0.0104, -0.0066, 0.0111, ..., -0.0004, 0.0092, 0.0520], + [-0.0063, 0.0132, 0.0398, ..., -0.0101, -0.0179, -0.0064], + ..., + [-0.0034, 0.0175, -0.0096, ..., 0.0225, 0.0045, 0.0066], + [-0.0209, -0.0081, 0.0086, ..., -0.0308, 0.0109, 0.0222], + [ 0.0325, 0.0166, -0.0221, ..., 0.0199, 0.0442, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img.weight': tensor([[ 0.0054, 0.0354, -0.0082, ..., 0.0117, 0.0356, -0.0186], + [ 0.0513, 0.0141, -0.0010, ..., 0.0376, 0.0532, 0.0026], + [ 0.0175, -0.0347, -0.0013, ..., -0.0063, 0.0057, -0.0283], + ..., + [ 0.0354, -0.0048, -0.0209, ..., -0.0153, 0.0262, -0.0113], + [ 0.0037, 0.0244, 0.0356, ..., -0.0026, 0.0170, 0.0117], + [ 0.0486, 0.0175, 0.0043, ..., 0.0016, -0.0033, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.0.mlp.layer1.weight': tensor([[-0.0013, 0.0019, -0.0005, ..., -0.0027, 0.0015, -0.0011], + [-0.0006, 0.0024, -0.0004, ..., 0.0019, 0.0020, -0.0028], + [-0.0014, 0.0020, 0.0019, ..., -0.0007, 0.0019, -0.0021], + ..., + [-0.0003, -0.0009, 0.0006, ..., 0.0002, -0.0043, 0.0038], + [-0.0029, 0.0084, -0.0039, ..., 0.0016, 0.0046, -0.0003], + [-0.0029, -0.0022, 0.0022, ..., 0.0043, -0.0025, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.0.mlp.layer2.weight': tensor([[ 1.3962e-03, 4.0894e-03, 1.9150e-03, ..., -1.5335e-03, + -4.3030e-03, -3.3569e-03], + [-1.6174e-03, -1.9989e-03, -2.4872e-03, ..., 1.0452e-03, + -5.0964e-03, 3.3264e-03], + [ 1.8005e-03, 1.7548e-03, -2.3956e-03, ..., -7.2002e-05, + -8.3447e-05, -5.3024e-04], + ..., + [ 2.0142e-03, 1.6479e-03, -1.6327e-03, ..., -8.6594e-04, + 3.7842e-03, -2.0905e-03], + [-2.0752e-03, 1.4400e-04, 4.3945e-03, ..., 2.1515e-03, + 1.1230e-02, -6.4392e-03], + [ 2.5787e-03, 1.7929e-03, 1.2054e-03, ..., -2.6703e-03, + 4.6158e-04, -1.7014e-03]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0459, -0.0449, -0.0437, ..., 0.0033, 0.0035, 0.0035], + [ 0.0488, 0.0479, 0.0466, ..., -0.0028, -0.0029, -0.0030], + [ 0.0479, 0.0471, 0.0461, ..., -0.0029, -0.0031, -0.0031], + ..., + [ 0.0447, 0.0439, 0.0425, ..., -0.0027, -0.0029, -0.0028], + [ 0.0413, 0.0403, 0.0391, ..., -0.0025, -0.0026, -0.0025], + [ 0.0461, 0.0449, 0.0442, ..., -0.0033, -0.0035, -0.0034]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0035, 0.0040, 0.0038, ..., 0.0048, 0.0049, 0.0036], + [-0.0013, 0.0010, 0.0010, ..., 0.0015, 0.0018, 0.0016], + [-0.0034, 0.0037, 0.0033, ..., 0.0050, 0.0057, 0.0033], + ..., + [-0.0601, 0.0698, 0.0684, ..., 0.0669, 0.0635, 0.0608], + [-0.0075, 0.0079, 0.0079, ..., 0.0063, 0.0050, 0.0078], + [-0.0001, -0.0006, -0.0005, ..., -0.0004, -0.0002, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.1.weight': tensor([[-1.9531e-03, -1.9531e-03, -1.9455e-03, ..., -4.7493e-04, + -4.0436e-04, -5.6839e-04], + [-2.8076e-02, -2.6855e-02, -2.6855e-02, ..., 2.7537e-05, + 3.5286e-05, 4.5776e-05], + [ 3.4912e-02, 3.3936e-02, 3.3447e-02, ..., -2.2984e-04, + -2.1267e-04, -2.3651e-04], + ..., + [ 2.7954e-02, 2.7344e-02, 2.6855e-02, ..., -1.1158e-04, + -4.3869e-05, -8.0585e-05], + [ 3.8330e-02, 3.7598e-02, 3.6865e-02, ..., -3.0136e-04, + -2.4128e-04, -2.9945e-04], + [ 3.1738e-02, 3.0396e-02, 2.9907e-02, ..., -1.5736e-04, + -7.9155e-05, -1.2970e-04]], dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0110, -0.0068, 0.0027, ..., 0.0068, 0.0079, 0.0012], + [-0.0028, -0.0003, -0.0011, ..., 0.0004, -0.0018, -0.0011], + [-0.0018, -0.0026, -0.0008, ..., 0.0026, 0.0017, -0.0017], + ..., + [ 0.0013, -0.0138, 0.0226, ..., 0.0147, 0.0293, 0.0168], + [ 0.0096, 0.0019, 0.0168, ..., -0.0016, 0.0090, 0.0170], + [-0.0123, -0.0019, -0.0042, ..., 0.0020, -0.0018, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.1.weight': tensor([[-0.0031, -0.0032, -0.0027, ..., 0.0013, 0.0016, 0.0015], + [ 0.0500, 0.0483, 0.0474, ..., -0.0024, -0.0028, -0.0028], + [ 0.0425, 0.0415, 0.0408, ..., -0.0038, -0.0041, -0.0041], + ..., + [ 0.0820, 0.0796, 0.0786, ..., -0.0009, -0.0010, -0.0011], + [ 0.0024, 0.0024, 0.0024, ..., -0.0013, -0.0014, -0.0013], + [ 0.0547, 0.0535, 0.0520, ..., -0.0011, -0.0011, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.0.adaln_modulation_mlp.2.weight': tensor([[ 0.0005, 0.0078, 0.0083, ..., 0.0083, -0.0005, 0.0057], + [-0.0012, 0.0008, 0.0017, ..., 0.0039, 0.0007, -0.0002], + [ 0.0018, 0.0023, 0.0028, ..., 0.0023, -0.0019, 0.0027], + ..., + [ 0.0023, 0.0371, 0.0608, ..., 0.0273, -0.0019, 0.0129], + [ 0.0073, 0.0101, 0.0109, ..., 0.0143, -0.0071, 0.0098], + [-0.0083, -0.0038, -0.0037, ..., -0.0029, 0.0082, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_proj.weight': tensor([[-0.0253, 0.0165, -0.0099, ..., 0.0093, -0.0049, -0.0032], + [ 0.0106, -0.0068, -0.0005, ..., -0.0076, -0.0041, -0.0435], + [ 0.0148, -0.0056, -0.0059, ..., 0.0030, 0.0069, -0.0123], + ..., + [ 0.0018, 0.0031, 0.0046, ..., 0.0011, 0.0009, -0.0038], + [ 0.0042, -0.0052, -0.0023, ..., -0.0001, -0.0009, 0.0035], + [ 0.0040, -0.0033, 0.0010, ..., -0.0050, 0.0003, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm.weight': tensor([ 1.2988e-01, 4.1211e-01, 1.5625e-02, 5.4297e-01, 5.0781e-01, + 9.5695e-34, -1.9259e-34, -2.3499e-03, 1.2458e-33, 1.1368e-03, + 3.0093e-34, -1.5517e-35, 5.2963e-34, -1.2064e-04, 2.5117e-27, + 1.0954e-33, 1.0107e-01, 1.9184e-34, 1.9184e-34, -3.8605e-03, + 6.3941e-32, 3.5858e-03, 1.5137e-01, 7.1484e-01, 6.7578e-01, + 3.0078e-01, -5.0735e-04, 3.7891e-01, 1.3885e-03, 5.6391e-31, + 1.8120e-05, -1.0352e-33, 4.9472e-06, -5.0459e-32, -2.6894e-04, + 1.9379e-03, -7.3426e-34, -2.3472e-34, 1.2037e-34, 5.2060e-34, + -3.0845e-34, 4.7302e-03, 3.5509e-34, 7.9297e-01, 2.1875e-01, + 1.9238e-01, 1.8555e-01, 6.7188e-01, 3.7384e-04, -1.1635e-13, + -4.2021e-22, 1.5527e-01, 1.6394e-30, 9.9182e-04, 8.1250e-34, + -9.2983e-05, 1.7090e-03, 4.5776e-05, -1.9259e-34, -1.0114e-09, + 5.8982e-33, 2.3713e-33, 1.3241e-34, 3.5553e-03, 2.7539e-01, + 1.3965e-01, 3.9844e-01, 1.0205e-01, 1.4465e-02, 2.2269e-33, + 1.3843e-34, -5.0964e-03, -2.2870e-34, -2.1820e-03, -9.7049e-35, + 2.9942e-34, -6.9815e-33, 4.9133e-03, 1.3248e-25, 1.6467e-32, + 9.0027e-04, 9.3663e-35, 2.6963e-33, -2.3484e-05, -9.8102e-34, + 1.7822e-02, 7.3047e-01, 1.6504e-01, 1.3965e-01, 6.3672e-01, + 1.3086e-01, 9.5703e-02, 5.0354e-03, 1.4637e-31, 6.9618e-05, + -5.2662e-34, -4.9174e-06, -9.5526e-31, -3.2037e-06, 1.8692e-03, + -1.6927e-34, 8.1852e-34, 2.8588e-35, -2.4826e-34, 1.9410e-34, + -6.9141e-06, -8.1250e-35, 1.1914e-01, 7.2656e-01, 6.8750e-01, + 6.9531e-01, 8.4473e-02, 3.3398e-01, 1.0991e-14, 3.1102e-22, + 4.7302e-04, 2.8843e-30, 6.3477e-03, 3.6563e-34, -5.6458e-03, + -4.6692e-03, 3.2654e-03, -1.5347e-34, -3.6948e-12, -1.1170e-32, + 2.6602e-33, 2.5579e-34, 2.0447e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.k_proj.weight': tensor([[ 0.0040, -0.0139, 0.0041, ..., -0.0057, 0.0012, -0.0267], + [ 0.0270, -0.0109, -0.0023, ..., -0.0145, 0.0079, -0.0232], + [ 0.0115, 0.0116, 0.0043, ..., 0.0038, 0.0016, -0.0044], + ..., + [-0.0043, 0.0004, -0.0066, ..., -0.0013, 0.0016, -0.0023], + [ 0.0041, -0.0035, 0.0007, ..., 0.0028, 0.0049, 0.0018], + [ 0.0041, 0.0005, -0.0008, ..., 0.0021, -0.0031, 0.0022]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm.weight': tensor([ 1.7871e-01, 4.2969e-01, 4.6692e-03, 5.5859e-01, 5.1172e-01, + 4.7546e-34, 2.8889e-34, -3.4180e-03, 3.9120e-34, -1.0834e-03, + 2.7685e-34, 2.9340e-34, -1.1977e-33, -1.6689e-04, -2.5370e-27, + -1.1194e-33, 1.0107e-01, 2.4375e-34, -6.4699e-34, 2.6733e-02, + -4.9304e-32, 3.6011e-03, 1.6895e-01, 7.2266e-01, 6.7969e-01, + 5.6152e-02, 1.0400e-01, 4.2773e-01, 2.4414e-03, -9.2445e-31, + 1.6212e-05, -1.6626e-34, 6.0201e-06, -4.6030e-32, -2.7466e-04, + 2.1484e-02, 1.6250e-33, -2.1667e-34, 9.5544e-35, -4.3935e-34, + -1.9259e-34, 4.7302e-03, -2.9039e-34, 7.8516e-01, 1.8652e-01, + -2.4261e-03, -3.2349e-03, 6.6797e-01, 3.2806e-03, 7.0166e-14, + 3.2591e-22, 7.7637e-02, 2.1940e-30, 4.5776e-04, -3.8519e-34, + 1.3351e-04, 1.7014e-03, -8.1539e-05, 2.3021e-34, -9.8953e-10, + -1.7935e-33, -4.3333e-34, 5.3791e-35, 3.5706e-03, 2.7344e-01, + 1.4453e-01, 4.0234e-01, 6.1768e-02, 3.9062e-03, -2.9039e-34, + -1.1586e-34, -3.6469e-03, -3.6111e-34, -2.2736e-03, -1.2413e-34, + 1.3466e-34, -5.5370e-33, 5.0049e-03, 1.3248e-25, -1.6756e-32, + -2.0752e-03, -1.6833e-35, -2.2389e-33, -1.5199e-05, -1.8417e-33, + 1.7944e-02, 7.2656e-01, 4.7363e-02, 4.8828e-02, 6.4062e-01, + 1.2354e-01, 1.4941e-01, -4.6692e-03, 2.2187e-31, -8.6784e-05, + -9.0880e-34, 3.7104e-06, 9.9840e-31, 3.7432e-05, 3.1250e-02, + 2.0463e-34, -1.0232e-33, 3.1748e-34, 3.2951e-34, 9.5920e-35, + -7.9274e-06, -1.1435e-34, 1.8457e-01, 7.3047e-01, 6.9141e-01, + 6.9531e-01, -1.6861e-03, 3.4961e-01, 7.6827e-14, 4.7315e-22, + -1.2500e-01, 4.1292e-31, 6.5002e-03, 2.9190e-34, 5.7373e-03, + -5.0049e-03, 3.2349e-03, -2.7154e-36, 3.6948e-12, 1.0015e-32, + -1.5648e-33, 1.4219e-34, 2.0142e-03], dtype=torch.bfloat16), 'net.blocks.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.self_attn.v_proj.weight': tensor([[ 0.0114, 0.0034, 0.0112, ..., -0.0095, 0.0038, 0.0052], + [ 0.0087, 0.0016, 0.0084, ..., 0.0074, 0.0022, -0.0032], + [ 0.0102, 0.0066, -0.0027, ..., 0.0016, 0.0046, -0.0096], + ..., + [ 0.0098, 0.0017, 0.0097, ..., 0.0034, 0.0007, -0.0127], + [-0.0022, -0.0018, -0.0012, ..., 0.0369, -0.0015, 0.0025], + [ 0.0098, -0.0025, -0.0025, ..., 0.0073, -0.0027, -0.0025]], + dtype=torch.bfloat16), 'net.blocks.1.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0049, -0.0111, ..., 0.0071, -0.0193, 0.0027], + [ 0.0142, 0.0031, 0.0162, ..., 0.0062, -0.0090, 0.0138], + [-0.0018, 0.0032, -0.0011, ..., 0.0017, 0.0037, 0.0024], + ..., + [-0.0019, 0.0006, -0.0024, ..., -0.0062, -0.0349, -0.0140], + [ 0.0019, -0.0029, -0.0046, ..., -0.0070, 0.0038, -0.0054], + [ 0.0021, 0.0004, 0.0024, ..., 0.0095, 0.0004, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_proj.weight': tensor([[-1.2024e-02, 1.2390e-02, 7.5073e-03, ..., -6.8359e-03, + 4.3335e-03, 3.0518e-03], + [ 5.4321e-03, -2.3603e-05, 3.4027e-03, ..., 2.7771e-03, + -4.0588e-03, -2.3193e-02], + [ 6.7139e-03, 9.2773e-03, -5.0735e-04, ..., -3.7537e-03, + -2.9907e-03, -6.4697e-03], + ..., + [-4.9133e-03, 5.9891e-04, -2.5330e-03, ..., 4.6997e-03, + 4.1504e-03, 7.2021e-03], + [-9.2773e-03, 2.8381e-03, 1.9836e-03, ..., 4.2725e-03, + 1.8463e-03, 1.3306e-02], + [ 2.2461e-02, -1.8158e-03, -2.0599e-03, ..., 3.0212e-03, + 2.4872e-03, -2.0874e-02]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm.weight': tensor([ 2.0996e-02, 4.3164e-01, 3.2471e-02, 4.7607e-03, 2.8839e-03, + 7.6294e-05, 4.6997e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.0771e-02, 2.3438e-02, 1.0147e-03, 1.7456e-02, 2.1851e-02, + 1.6357e-02, 1.2451e-02, -6.6757e-05, 3.0398e-06, 2.9053e-02, + 2.3828e-01, 2.3804e-02, 8.3618e-03, 2.1729e-02, 8.0566e-03, + 4.3297e-04, 6.5193e-07, 7.1777e-02, 4.2282e-07, 3.3789e-01, + -1.4976e-06, -1.7360e-06, 1.1780e-02, 2.6107e-05, 1.0681e-02, + 1.4221e-02, 4.4405e-06, 3.6523e-01, 1.1780e-02, 3.0823e-03, + 8.1543e-02, 7.4158e-03, 1.6479e-02, 6.0730e-03, 6.6223e-03, + 1.0376e-02, 8.3618e-03, 1.9531e-02, 2.7588e-02, 1.5076e-02, + 1.1841e-02, 3.2227e-01, 1.1963e-02, 2.3193e-02, 3.5400e-02, + 4.3030e-03, 9.8145e-02, 1.3794e-02, 2.7954e-02, 2.0020e-02, + 1.1742e-05, 1.6357e-02, 1.2146e-02, -1.2457e-05, 1.5869e-02, + 1.0803e-02, 4.6194e-07, 1.0376e-03, 1.2024e-02, 1.0400e-01, + 1.0452e-03, 9.2163e-03, 7.2754e-02, 4.6143e-02, 5.7983e-03, + 1.2451e-02, -2.7061e-05, 2.9419e-02, 2.6733e-02, 3.1128e-02, + 3.7384e-03, 2.1362e-02, 3.7598e-02, 4.0054e-05, 3.0640e-02, + 1.0132e-02, 3.2349e-03, 5.5908e-02, 1.6846e-02, 1.7700e-03, + 1.3916e-02, 4.3640e-03, 1.2939e-02, 1.8066e-02, 9.8267e-03, + 1.0109e-04, 1.6724e-02, 4.5410e-02, 1.5503e-02, 5.4199e-02, + 4.2725e-02, 3.9307e-02, 1.3062e-02, 1.5974e-05, -9.2316e-04, + 1.3000e-02, 5.8594e-03, 2.3804e-02, 1.9165e-02, 2.7832e-02, + 4.0894e-03, 1.0071e-02, 4.6875e-02, 4.5508e-01, 4.7607e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7212e-02, 1.0547e-01, + 1.4572e-03, 3.1090e-04, 1.8311e-04, 1.7456e-02, 1.1292e-02, + 2.0874e-02, 3.3203e-02, 1.9531e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_proj.weight': tensor([[ 8.3542e-04, -3.1586e-03, -3.2349e-03, ..., -3.6621e-03, + -5.0659e-03, 8.5831e-04], + [ 2.9907e-03, -3.3417e-03, 8.6670e-03, ..., 1.0437e-02, + 2.3315e-02, 3.5858e-03], + [ 1.2512e-03, 3.4943e-03, 3.7384e-03, ..., 4.6692e-03, + 4.7913e-03, -4.3640e-03], + ..., + [-1.4648e-03, -4.7607e-03, -7.4387e-04, ..., 4.5776e-05, + -3.9062e-03, 2.2278e-03], + [-3.0518e-03, -5.1880e-03, -1.1963e-02, ..., 4.1389e-04, + 1.5030e-03, 3.6774e-03], + [ 5.0049e-03, -2.7084e-04, -8.7738e-04, ..., 4.5471e-03, + -1.0986e-03, 5.7678e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm.weight': tensor([ 2.1362e-02, 4.3164e-01, 3.2227e-02, 4.8523e-03, 2.8839e-03, + 7.5817e-05, 4.7302e-03, 5.1172e-01, 2.7771e-03, 3.3984e-01, + 4.1504e-02, 2.3438e-02, 1.0376e-03, 1.7456e-02, 2.1729e-02, + 1.6235e-02, 1.2695e-02, -6.6280e-05, 3.0547e-06, 2.9053e-02, + 2.3828e-01, 2.3315e-02, 8.3618e-03, 2.1973e-02, 8.0566e-03, + 4.2725e-04, 5.1409e-07, 7.1289e-02, -4.8801e-07, 3.3789e-01, + 9.7603e-07, -1.4007e-06, 1.1658e-02, 1.8835e-05, 1.0803e-02, + 1.4221e-02, 4.5300e-06, 3.6523e-01, 1.1841e-02, 3.0060e-03, + 8.0566e-02, 7.3547e-03, 1.6235e-02, 6.1035e-03, 6.5002e-03, + 1.0376e-02, 8.4839e-03, 1.9409e-02, 2.7588e-02, 1.5320e-02, + 1.1780e-02, 3.2227e-01, 1.1963e-02, 2.2949e-02, 3.5156e-02, + 4.2419e-03, 9.8145e-02, 1.3733e-02, 2.7588e-02, 2.0142e-02, + 1.1921e-05, 1.6235e-02, 1.2207e-02, -1.2398e-05, 1.5869e-02, + 1.0742e-02, 3.1106e-07, 1.0529e-03, 1.2146e-02, 1.0400e-01, + 9.5367e-04, 9.3384e-03, 7.2754e-02, 4.6143e-02, 5.8289e-03, + 1.2695e-02, -2.7895e-05, 2.9297e-02, 2.6733e-02, 3.1128e-02, + 3.7689e-03, 2.1362e-02, 3.6865e-02, 4.0054e-05, 3.0762e-02, + 1.0132e-02, 3.1738e-03, 5.5420e-02, 1.7090e-02, 1.9684e-03, + 1.4221e-02, 4.3640e-03, 1.3000e-02, 1.7944e-02, 9.7046e-03, + 9.9182e-05, 1.6235e-02, 4.5166e-02, 1.5503e-02, 5.3955e-02, + 4.3457e-02, 3.9795e-02, 1.3123e-02, -1.4976e-06, 9.3460e-04, + 1.3000e-02, 5.9204e-03, 2.3804e-02, 1.9287e-02, 2.7710e-02, + 4.1809e-03, 9.7046e-03, 4.6387e-02, 4.5508e-01, 4.7913e-03, + 2.7466e-02, 1.2878e-02, 1.4465e-02, 1.7334e-02, 1.0547e-01, + 1.6632e-03, 2.8992e-04, 1.8406e-04, 1.7822e-02, 1.1292e-02, + 2.1118e-02, 3.3203e-02, 1.9409e-02], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.v_proj.weight': tensor([[-1.6602e-02, -3.1710e-05, -3.3760e-04, ..., 5.6839e-04, + -2.1973e-02, -2.3682e-02], + [-6.2256e-03, 4.6387e-03, 3.4790e-03, ..., 1.6174e-03, + 8.4229e-03, 1.0071e-02], + [-3.8330e-02, -2.0905e-03, -3.2654e-03, ..., -6.8970e-03, + -3.0273e-02, -9.5215e-03], + ..., + [-4.1992e-02, 1.8463e-03, -1.7944e-02, ..., -2.2316e-04, + 1.8677e-02, 8.1177e-03], + [ 1.1841e-02, 6.2561e-03, -1.8005e-03, ..., 3.7231e-03, + 1.0742e-02, 4.3030e-03], + [ 9.2773e-03, 3.8452e-03, -8.1177e-03, ..., 2.0142e-03, + 8.0566e-03, 2.7313e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.output_proj.weight': tensor([[-2.8839e-03, -2.0905e-03, 2.0996e-02, ..., -2.5940e-03, + 7.9346e-03, -8.4686e-04], + [-6.5613e-03, -7.1716e-03, -5.6152e-03, ..., -7.3547e-03, + -9.7656e-04, 1.4465e-02], + [-7.1526e-05, 2.5787e-03, -3.4637e-03, ..., 1.7853e-03, + -8.4686e-04, -2.3956e-03], + ..., + [ 2.6703e-04, 8.1635e-04, -7.9727e-04, ..., -7.3242e-04, + -2.7313e-03, 2.3041e-03], + [ 3.8757e-03, 5.7373e-03, 4.8523e-03, ..., -3.2959e-02, + 2.1057e-03, 2.9541e-02], + [ 4.2419e-03, 3.1738e-03, -2.0996e-02, ..., -1.2817e-02, + 1.0681e-02, 3.8605e-03]], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img.weight': tensor([[ 0.0142, 0.0569, 0.0117, ..., -0.0075, -0.0337, 0.0076], + [ 0.0491, 0.0413, 0.0251, ..., -0.0060, 0.0393, 0.0040], + [-0.0162, 0.0039, 0.0135, ..., -0.0194, -0.0247, 0.0469], + ..., + [-0.0026, 0.0261, 0.0041, ..., -0.0038, 0.0347, 0.0053], + [-0.0254, 0.0068, -0.0012, ..., 0.0096, 0.0388, -0.0200], + [ 0.0240, 0.0190, 0.0284, ..., -0.0491, 0.0124, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.v_img.weight': tensor([[ 0.0033, -0.0270, -0.0166, ..., 0.0117, -0.0298, 0.0532], + [-0.0016, 0.0044, -0.0129, ..., -0.0161, 0.0150, 0.0308], + [ 0.0137, 0.0084, 0.0139, ..., -0.0120, -0.0410, 0.0354], + ..., + [ 0.0053, -0.0630, 0.0005, ..., 0.0074, -0.0183, 0.0214], + [ 0.0034, -0.0018, 0.0258, ..., -0.0486, -0.0349, -0.0044], + [ 0.0182, -0.0074, 0.0198, ..., 0.0152, -0.0225, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img.weight': tensor([[ 0.0036, -0.0028, -0.0006, ..., 0.0035, 0.0022, -0.0117], + [-0.0154, 0.0251, 0.0432, ..., -0.0052, 0.0034, -0.0016], + [-0.0144, -0.0203, -0.0245, ..., -0.0112, -0.0325, -0.0018], + ..., + [ 0.0050, 0.0037, 0.0417, ..., 0.0281, 0.0025, 0.0030], + [-0.0038, -0.0046, 0.0261, ..., 0.0205, -0.0286, -0.0325], + [-0.0227, -0.0364, -0.0220, ..., 0.0062, -0.0039, -0.0091]], + dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.1.mlp.layer1.weight': tensor([[-8.8501e-03, -6.1035e-05, -9.2697e-04, ..., 3.8605e-03, + 2.3956e-03, -7.5378e-03], + [-2.9755e-03, 2.5024e-03, -3.8757e-03, ..., 3.0975e-03, + -8.7738e-04, -8.4229e-03], + [-2.3041e-03, 1.7929e-03, -1.9379e-03, ..., 2.8534e-03, + -1.3733e-03, -3.3188e-04], + ..., + [ 2.7832e-02, 9.1553e-04, -2.3193e-03, ..., -2.0752e-03, + -1.9989e-03, -3.8300e-03], + [-2.2217e-02, -1.6022e-04, -5.0964e-03, ..., 1.1841e-02, + -2.9907e-03, -2.1210e-03], + [ 1.6479e-02, -9.0027e-04, 2.3804e-03, ..., -1.6403e-03, + 1.1520e-03, -9.2163e-03]], dtype=torch.bfloat16), 'net.blocks.1.mlp.layer2.weight': tensor([[-0.0082, -0.0067, 0.0034, ..., -0.0024, 0.0007, -0.0057], + [-0.0001, 0.0049, 0.0097, ..., 0.0005, 0.0066, 0.0047], + [-0.0022, -0.0028, -0.0001, ..., -0.0041, 0.0009, -0.0022], + ..., + [ 0.0060, 0.0049, 0.0066, ..., 0.0024, -0.0117, 0.0192], + [-0.0026, 0.0092, 0.0026, ..., 0.0051, 0.0026, -0.0012], + [ 0.0054, -0.0017, -0.0027, ..., -0.0009, 0.0014, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0435, -0.0422, -0.0417, ..., 0.0053, 0.0057, 0.0056], + [ 0.0272, 0.0261, 0.0250, ..., -0.0041, -0.0040, -0.0041], + [-0.0461, -0.0442, -0.0437, ..., 0.0054, 0.0056, 0.0055], + ..., + [-0.0491, -0.0481, -0.0469, ..., 0.0037, 0.0039, 0.0040], + [-0.0432, -0.0427, -0.0415, ..., 0.0040, 0.0043, 0.0043], + [-0.0172, -0.0164, -0.0153, ..., 0.0015, 0.0016, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0129, 0.0006, 0.0091, ..., 0.0105, 0.0104, -0.0007], + [-0.0023, 0.0011, -0.0020, ..., -0.0017, -0.0015, -0.0008], + [-0.0060, -0.0008, -0.0053, ..., -0.0080, -0.0072, -0.0001], + ..., + [-0.0288, -0.0016, -0.0291, ..., -0.0325, -0.0308, 0.0022], + [-0.0136, -0.0073, -0.0131, ..., -0.0139, -0.0124, 0.0086], + [ 0.0022, -0.0089, 0.0037, ..., 0.0019, 0.0019, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.1.weight': tensor([[ 4.4922e-02, 4.3701e-02, 4.2725e-02, ..., 3.4273e-06, + -4.2915e-05, 5.1856e-06], + [ 7.8735e-03, 8.3008e-03, 8.0566e-03, ..., -2.1744e-04, + -2.9564e-04, -4.1008e-04], + [-8.7280e-03, -8.4229e-03, -7.8125e-03, ..., -5.1880e-04, + -5.1880e-04, -5.3024e-04], + ..., + [-2.1606e-02, -2.1729e-02, -2.1118e-02, ..., -5.6458e-04, + -5.2261e-04, -4.8828e-04], + [-2.9297e-02, -2.8687e-02, -2.7710e-02, ..., -3.7956e-04, + -3.3379e-04, -3.8719e-04], + [ 4.3701e-02, 4.2480e-02, 4.1748e-02, ..., -5.0545e-05, + 6.7234e-05, -7.9632e-05]], dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0101, -0.0037, -0.0118, ..., -0.0044, 0.0103, -0.0104], + [ 0.0013, 0.0036, -0.0020, ..., 0.0020, -0.0023, 0.0007], + [ 0.0066, 0.0018, 0.0002, ..., 0.0022, -0.0043, 0.0062], + ..., + [ 0.0796, 0.0104, -0.0046, ..., 0.0070, -0.0128, 0.0693], + [ 0.0038, -0.0071, -0.0044, ..., -0.0120, -0.0188, 0.0098], + [-0.0011, 0.0095, -0.0153, ..., 0.0104, 0.0066, -0.0028]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.1.weight': tensor([[-0.0311, -0.0308, -0.0304, ..., 0.0002, 0.0001, 0.0001], + [ 0.0500, 0.0493, 0.0488, ..., -0.0036, -0.0036, -0.0036], + [ 0.0483, 0.0474, 0.0466, ..., -0.0037, -0.0038, -0.0038], + ..., + [ 0.0527, 0.0513, 0.0508, ..., -0.0020, -0.0020, -0.0019], + [ 0.0181, 0.0183, 0.0182, ..., 0.0016, 0.0016, 0.0015], + [ 0.0469, 0.0466, 0.0461, ..., -0.0023, -0.0024, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.1.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0002, -0.0006, ..., 0.0002, -0.0008, -0.0008], + [ 0.0020, 0.0005, 0.0014, ..., 0.0006, -0.0024, -0.0002], + [ 0.0031, -0.0062, -0.0071, ..., -0.0051, -0.0025, -0.0053], + ..., + [-0.0034, -0.0065, -0.0114, ..., -0.0046, 0.0054, -0.0036], + [-0.0034, 0.0034, 0.0027, ..., 0.0033, 0.0034, 0.0034], + [ 0.0062, -0.0020, 0.0033, ..., -0.0020, -0.0068, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_proj.weight': tensor([[ 0.0208, -0.0262, -0.0146, ..., 0.0046, -0.0148, 0.0405], + [-0.0098, -0.0100, -0.0052, ..., -0.0063, 0.0031, -0.0103], + [ 0.0107, 0.0086, 0.0127, ..., -0.0010, -0.0103, 0.0199], + ..., + [-0.0151, -0.0074, -0.0031, ..., 0.0033, 0.0019, -0.0035], + [ 0.0215, 0.0248, 0.0013, ..., 0.0022, -0.0038, 0.0075], + [ 0.0018, 0.0026, 0.0166, ..., -0.0219, 0.0027, -0.0092]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm.weight': tensor([ 1.0859e+00, 9.9121e-02, 2.8516e-01, 2.5391e-01, 9.1406e-01, + 5.7678e-03, 6.4087e-03, -1.3638e-04, 1.3123e-02, 5.3101e-03, + -1.1730e-04, -9.7752e-05, 8.6719e-01, -5.2261e-04, 1.9222e-06, + -3.5095e-03, -1.3123e-02, -4.3030e-03, 1.6602e-02, 1.8066e-02, + 1.5820e-01, 1.2054e-03, 1.9238e-01, 1.1406e+00, 5.4297e-01, + 5.2344e-01, 2.3535e-01, 9.6191e-02, 2.5757e-02, -5.6152e-03, + 8.1787e-03, 2.6131e-04, 8.0566e-03, -1.8387e-03, 9.7275e-04, + 3.4180e-02, 1.2402e-01, 2.2888e-04, 9.6893e-04, -2.2507e-04, + 3.1948e-05, 9.6321e-05, 6.7520e-04, 9.3359e-01, 1.4587e-02, + 1.0391e+00, 5.6641e-01, 1.1328e-01, -3.9062e-02, 5.3406e-03, + -6.6528e-03, 2.7954e-02, 1.3086e-01, 7.2420e-06, -1.6499e-04, + -7.4863e-05, -1.9907e-08, 1.6212e-04, -5.7617e-02, 1.0538e-04, + 3.1250e-02, -2.5034e-05, -1.7212e-02, 2.3956e-03, 3.8477e-01, + 1.0234e+00, 9.6484e-01, 9.4531e-01, 1.8262e-01, -4.8828e-03, + 1.2817e-02, -1.5545e-04, 2.7222e-02, -2.6131e-04, -1.1492e-04, + 7.8678e-05, 4.2969e-02, 8.3008e-03, 1.1504e-05, 5.4688e-01, + 8.2422e-01, -4.1199e-03, 3.4904e-04, 3.4180e-02, 1.1328e-01, + 1.4648e-02, 1.1172e+00, 1.6211e-01, 9.8047e-01, 1.0312e+00, + 1.4160e-01, 3.2812e-01, 3.7598e-02, -5.9814e-03, -3.7994e-03, + -1.2684e-04, -1.6113e-02, -3.1281e-03, 5.0049e-03, 3.9062e-02, + 1.1719e-01, 9.0122e-05, 8.7891e-01, -1.8692e-04, -7.4387e-05, + -2.8229e-04, 8.9169e-05, 6.3281e-01, 1.1875e+00, 1.0469e+00, + 1.1094e+00, 2.8711e-01, 8.2397e-03, 8.4229e-03, 4.7302e-03, + 2.4658e-02, 1.3965e-01, -1.2302e-04, -4.5538e-05, 2.9325e-05, + 1.4114e-04, -2.8038e-04, 9.1797e-01, -6.0558e-05, 2.4902e-02, + 3.3379e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.k_proj.weight': tensor([[-0.0206, 0.0111, 0.0132, ..., 0.0145, 0.0019, 0.0165], + [ 0.0093, 0.0057, -0.0014, ..., 0.0026, 0.0078, -0.0071], + [-0.0254, 0.0134, -0.0094, ..., 0.0165, 0.0220, 0.0027], + ..., + [ 0.0011, 0.0028, 0.0009, ..., -0.0035, -0.0017, 0.0031], + [-0.0048, -0.0050, 0.0083, ..., 0.0059, -0.0108, 0.0018], + [ 0.0009, -0.0088, 0.0025, ..., -0.0144, 0.0040, 0.0203]], + dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm.weight': tensor([ 1.1094e+00, 5.4297e-01, 2.3047e-01, 1.8652e-01, 9.1797e-01, + 6.7139e-03, 1.5747e-02, -1.2779e-04, 1.4954e-02, 5.1880e-03, + -1.2016e-04, 9.5367e-05, 8.6719e-01, 4.7493e-04, 4.8801e-07, + -1.1414e-02, -2.8229e-03, 8.2031e-02, 1.8188e-02, -5.3406e-03, + 2.2736e-03, 1.1826e-03, 3.9453e-01, 1.0156e+00, 5.4443e-02, + 4.1406e-01, 2.8320e-01, 1.1865e-01, -1.9379e-03, 6.6528e-03, + 8.1787e-03, -2.3651e-04, 7.7209e-03, 4.4434e-02, 1.5564e-03, + -3.8605e-03, 6.0730e-03, 2.4414e-04, 1.7471e-03, -2.3079e-04, + 9.1195e-06, 9.6321e-05, 6.7139e-04, 1.0625e+00, 6.1328e-01, + 1.0000e+00, 3.7305e-01, 1.9531e-01, 3.1433e-03, 5.9326e-02, + 7.9346e-03, -4.3030e-03, -3.8452e-03, -1.3649e-05, 1.5545e-04, + 9.3460e-05, -7.5340e-05, 1.6689e-04, -1.0620e-02, -1.0538e-04, + 1.9379e-03, 2.3723e-05, -1.8921e-02, -4.1199e-03, 3.6914e-01, + 9.4141e-01, 9.6875e-01, 9.3359e-01, 1.8457e-01, -4.6349e-04, + 5.5237e-03, 1.9550e-04, 2.2339e-02, -6.4087e-04, -1.1253e-04, + 7.6294e-05, -2.2736e-03, 8.4229e-03, -1.4186e-05, 5.4688e-01, + 8.2422e-01, 6.2988e-02, 7.3242e-04, -5.2185e-03, -3.8452e-03, + 1.4954e-02, 1.0391e+00, 6.7188e-01, 1.1016e+00, 1.0781e+00, + 1.4343e-02, 3.3008e-01, -2.8229e-03, 4.2419e-03, 5.7678e-03, + -1.4782e-04, 1.8311e-02, 4.6875e-02, 1.1292e-02, -1.0742e-02, + 4.0588e-03, -8.5354e-05, 8.8281e-01, 1.8311e-04, -6.0558e-05, + -2.8038e-04, -9.0122e-05, 2.0508e-01, 1.0547e+00, 1.0234e+00, + 1.2344e+00, 2.8906e-01, -3.1982e-02, 8.9355e-02, 2.6093e-03, + 8.4839e-03, 4.2725e-03, -1.2398e-04, -9.7752e-05, 7.4208e-06, + 1.1826e-04, -2.7084e-04, 9.1797e-01, -6.0797e-05, -4.3030e-03, + -3.2187e-05, 4.8438e-01, 9.0625e-01], dtype=torch.bfloat16), 'net.blocks.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.self_attn.v_proj.weight': tensor([[-2.0599e-03, 1.1169e-02, -1.1292e-02, ..., -2.0752e-02, + 5.8899e-03, 1.1292e-02], + [ 1.3885e-03, -4.2725e-03, -2.0874e-02, ..., -8.4229e-03, + 2.2888e-05, -1.5793e-03], + [-1.5381e-02, -1.5640e-03, 3.8452e-03, ..., 2.3438e-02, + -1.0254e-02, -1.2512e-02], + ..., + [ 8.9722e-03, 1.9684e-03, -3.0029e-02, ..., -4.1504e-02, + -8.3160e-04, -2.0409e-04], + [-1.1587e-04, 5.7602e-04, -3.1738e-03, ..., 1.9287e-02, + -2.6093e-03, -9.2163e-03], + [-5.2490e-03, -8.3008e-03, -2.3438e-02, ..., -2.9449e-03, + 8.0566e-03, -1.7700e-03]], dtype=torch.bfloat16), 'net.blocks.2.self_attn.output_proj.weight': tensor([[ 0.0069, 0.0261, -0.0410, ..., -0.0009, -0.0193, 0.0121], + [-0.0038, -0.0193, -0.0201, ..., 0.0059, 0.0150, -0.0266], + [-0.0017, -0.0018, -0.0027, ..., -0.0035, -0.0029, 0.0083], + ..., + [ 0.0086, 0.0076, -0.0183, ..., 0.0160, 0.0073, 0.0017], + [ 0.0014, -0.0018, 0.0044, ..., 0.0087, 0.0143, -0.0067], + [-0.0018, 0.0024, -0.0084, ..., -0.0038, 0.0078, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_proj.weight': tensor([[-0.0023, 0.0060, -0.0033, ..., -0.0010, 0.0065, 0.0005], + [ 0.0011, 0.0070, -0.0016, ..., -0.0007, -0.0029, 0.0078], + [ 0.0032, -0.0008, -0.0047, ..., -0.0034, 0.0137, -0.0038], + ..., + [ 0.0024, 0.0050, -0.0019, ..., 0.0101, -0.0045, -0.0019], + [ 0.0042, -0.0025, 0.0015, ..., 0.0126, -0.0003, 0.0058], + [ 0.0004, 0.0065, -0.0016, ..., -0.0097, 0.0024, -0.0039]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm.weight': tensor([1.9684e-03, 1.3574e-01, 2.0630e-02, 5.5859e-01, 1.9409e-02, 2.0020e-02, + 1.5869e-02, 9.5215e-03, 1.9165e-02, 1.2109e-01, 2.0142e-02, 5.9326e-02, + 2.5391e-01, 2.2949e-02, 1.8164e-01, 2.1484e-02, 6.4941e-02, 2.2461e-02, + 1.5030e-03, 1.5869e-02, 1.6602e-02, 1.7676e-01, 2.6758e-01, 1.9629e-01, + 2.5635e-02, 1.6211e-01, 6.7871e-02, 1.5564e-02, 3.5547e-01, 1.5747e-02, + 1.6846e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6357e-02, 1.5015e-02, + 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, 7.9346e-03, 1.1377e-01, + 1.1841e-02, 5.1562e-01, 3.5645e-02, 4.1992e-02, 8.3618e-03, 3.5938e-01, + 6.6223e-03, 1.7212e-02, 8.3008e-02, 1.3428e-02, 1.6235e-02, 4.2383e-01, + 2.8229e-03, 3.2227e-02, 6.0547e-02, 8.2520e-02, 7.1106e-03, 1.4038e-02, + 2.7771e-03, 2.0386e-02, 5.4321e-03, 3.2617e-01, 2.1118e-02, 3.8757e-03, + 3.1494e-02, 7.5684e-02, 7.7637e-02, 1.4893e-02, 6.3171e-03, 5.9766e-01, + 8.5354e-05, 5.4688e-01, 1.4038e-02, 1.0834e-03, 1.8921e-02, 2.8125e-01, + 3.2471e-02, 3.6926e-03, 1.0547e-01, 7.7515e-03, 5.9082e-02, 9.0942e-03, + 2.1362e-02, 1.3062e-02, 1.2390e-02, 7.4219e-02, 2.0312e-01, 6.8359e-03, + 2.6855e-03, 2.0630e-02, 5.2979e-02, 2.5757e-02, 3.1006e-02, 1.8555e-02, + 1.8311e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, 5.0293e-02, 1.6235e-02, + 1.4191e-03, 6.4453e-02, 5.5859e-01, 7.7637e-02, 7.9346e-03, 4.0039e-01, + 1.4771e-02, 3.0640e-02, 1.8433e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, + 3.7305e-01, 2.4536e-02, 3.2959e-02, 4.5703e-01, 1.2256e-01, 4.2480e-02, + 1.1047e-02, 6.6797e-01, 1.4746e-01, 2.1973e-02, 7.2021e-03, 2.5177e-03, + 2.7344e-02, 4.6692e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_proj.weight': tensor([[ 0.0025, -0.0032, 0.0097, ..., -0.0014, -0.0043, 0.0034], + [-0.0048, 0.0001, 0.0006, ..., 0.0060, -0.0063, -0.0098], + [-0.0068, -0.0018, 0.0004, ..., 0.0128, -0.0156, -0.0038], + ..., + [ 0.0015, 0.0048, -0.0068, ..., 0.0027, 0.0082, -0.0008], + [ 0.0126, 0.0028, 0.0042, ..., -0.0036, 0.0034, 0.0042], + [-0.0026, 0.0008, 0.0073, ..., -0.0030, -0.0109, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm.weight': tensor([ 1.9531e-03, 1.3574e-01, 2.0508e-02, 5.5859e-01, 1.9043e-02, + 1.9775e-02, 1.5991e-02, 9.4604e-03, 1.9287e-02, 1.2109e-01, + 2.0142e-02, 5.7373e-02, 2.5391e-01, 2.2827e-02, 1.8164e-01, + 2.2095e-02, 6.4941e-02, 2.2461e-02, 1.4038e-03, 1.5564e-02, + 1.6846e-02, 1.7773e-01, 2.6758e-01, 1.9629e-01, 2.5391e-02, + 1.6211e-01, 6.6895e-02, 1.3733e-02, 3.5742e-01, 1.5747e-02, + 1.7090e-02, 4.9023e-01, 6.9336e-02, 3.3936e-02, 1.6602e-02, + 1.4404e-02, 9.0332e-02, 4.0234e-01, 4.2969e-01, 1.1292e-02, + 7.8735e-03, 1.1377e-01, 1.1597e-02, 5.1953e-01, 3.6133e-02, + 4.1992e-02, 8.6670e-03, 3.5938e-01, 6.7749e-03, 1.6968e-02, + 8.3008e-02, 1.3733e-02, 1.6235e-02, 4.2383e-01, 3.1586e-03, + 3.2227e-02, 6.0059e-02, 8.2520e-02, 7.1411e-03, 1.2390e-02, + 2.6245e-03, 2.0264e-02, 5.7983e-03, 3.2617e-01, 2.0874e-02, + 3.9978e-03, 3.1006e-02, 7.5684e-02, 7.7637e-02, 1.4954e-02, + 6.7139e-03, 5.9766e-01, 8.7619e-06, 5.4688e-01, 1.4221e-02, + -6.3324e-04, 1.8921e-02, 2.8125e-01, 3.3203e-02, 3.8452e-03, + 1.0596e-01, 7.5073e-03, 5.8350e-02, 8.9111e-03, 2.0874e-02, + 1.3123e-02, 1.2939e-02, 7.4707e-02, 2.0312e-01, 6.6833e-03, + 2.7466e-03, 1.9531e-02, 5.3223e-02, 2.5635e-02, 3.1494e-02, + 1.8677e-02, 1.8433e-02, 2.6123e-02, 7.0801e-02, 2.7100e-02, + 4.9805e-02, 1.6357e-02, 2.4414e-03, 6.4941e-02, 5.5859e-01, + 7.7637e-02, 7.8735e-03, 4.0039e-01, 1.4404e-02, 3.0640e-02, + 1.9043e-02, 8.1787e-03, 2.2411e-04, 9.9121e-02, 3.7305e-01, + 2.4048e-02, 3.2959e-02, 4.5703e-01, 1.2354e-01, 4.2725e-02, + 1.0803e-02, 6.7188e-01, 1.4648e-01, 2.1118e-02, 6.8665e-03, + 2.3499e-03, 2.7100e-02, 4.4556e-03], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.v_proj.weight': tensor([[ 0.0122, -0.0050, -0.0166, ..., 0.0047, -0.0106, -0.0087], + [-0.0137, 0.0011, -0.0015, ..., 0.0001, -0.0251, 0.0182], + [-0.0144, -0.0037, 0.0049, ..., 0.0003, 0.0242, 0.0332], + ..., + [ 0.0053, -0.0045, -0.0140, ..., -0.0013, 0.0103, -0.0079], + [ 0.0201, 0.0053, 0.0219, ..., 0.0038, 0.0045, 0.0074], + [ 0.0220, 0.0049, 0.0058, ..., 0.0067, -0.0168, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.output_proj.weight': tensor([[ 0.0216, -0.0029, -0.0120, ..., -0.0159, 0.0153, 0.0043], + [ 0.0125, 0.0081, 0.0083, ..., 0.0193, -0.0171, 0.0056], + [ 0.0003, -0.0004, 0.0010, ..., -0.0036, 0.0007, -0.0052], + ..., + [-0.0006, -0.0029, 0.0018, ..., 0.0101, -0.0089, 0.0090], + [ 0.0164, -0.0205, 0.0044, ..., -0.0249, -0.0133, -0.0143], + [-0.0018, -0.0107, 0.0139, ..., 0.0302, 0.0045, 0.0146]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img.weight': tensor([[ 0.0156, -0.0078, -0.0481, ..., 0.0449, 0.0212, 0.0127], + [ 0.0154, -0.0227, 0.0220, ..., -0.0221, -0.0020, 0.0086], + [-0.0151, -0.0095, 0.0430, ..., -0.0101, 0.0327, -0.0112], + ..., + [ 0.0310, -0.0376, -0.0435, ..., -0.0225, 0.0255, 0.0146], + [-0.0025, 0.0166, -0.0063, ..., 0.0002, 0.0081, -0.0078], + [-0.0112, -0.0015, 0.0123, ..., 0.0042, -0.0036, -0.0082]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.v_img.weight': tensor([[-0.0025, -0.0386, -0.0123, ..., 0.0496, 0.0178, 0.0413], + [-0.0400, 0.0084, 0.0286, ..., -0.0339, -0.0160, 0.0166], + [-0.0131, 0.0019, -0.0124, ..., 0.0214, 0.0105, 0.0050], + ..., + [-0.0153, -0.0172, -0.0140, ..., 0.0095, 0.0101, 0.0184], + [-0.0154, -0.0356, 0.0201, ..., -0.0047, -0.0114, 0.0332], + [-0.0396, -0.0119, -0.0092, ..., -0.0284, -0.0142, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img.weight': tensor([[ 0.0276, 0.0312, 0.0136, ..., -0.0195, 0.0112, -0.0142], + [ 0.0192, 0.0081, -0.0077, ..., 0.0018, -0.0135, -0.0198], + [-0.0248, 0.0100, -0.0139, ..., 0.0090, 0.0172, 0.0061], + ..., + [ 0.0466, -0.0047, 0.0003, ..., -0.0008, 0.0239, 0.0047], + [ 0.0547, -0.0194, -0.0106, ..., -0.0017, 0.0311, -0.0044], + [ 0.0035, 0.0101, -0.0098, ..., 0.0203, 0.0170, 0.0143]], + dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.2.mlp.layer1.weight': tensor([[ 0.0011, -0.0043, -0.0011, ..., 0.0005, -0.0010, 0.0018], + [-0.0038, -0.0021, 0.0006, ..., -0.0007, -0.0037, 0.0021], + [ 0.0004, -0.0074, 0.0038, ..., -0.0078, 0.0004, -0.0515], + ..., + [-0.0018, 0.0010, -0.0023, ..., -0.0022, 0.0022, -0.0038], + [-0.0028, -0.0009, -0.0095, ..., -0.0121, -0.0052, -0.0281], + [ 0.0093, 0.0236, -0.0018, ..., 0.0054, -0.0018, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.2.mlp.layer2.weight': tensor([[-0.0027, 0.0010, -0.0109, ..., -0.0044, -0.0031, 0.0021], + [-0.0075, -0.0029, 0.0050, ..., 0.0044, -0.0128, -0.0065], + [-0.0009, -0.0010, 0.0015, ..., -0.0019, 0.0017, -0.0003], + ..., + [ 0.0023, 0.0002, -0.0130, ..., -0.0015, 0.0076, -0.0072], + [-0.0100, 0.0104, 0.0045, ..., 0.0036, 0.0004, -0.0056], + [ 0.0034, 0.0038, -0.0040, ..., -0.0067, 0.0037, -0.0413]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0378, 0.0374, 0.0361, ..., -0.0045, -0.0047, -0.0048], + [-0.0374, -0.0361, -0.0356, ..., 0.0047, 0.0052, 0.0052], + [-0.0376, -0.0371, -0.0361, ..., 0.0049, 0.0053, 0.0053], + ..., + [-0.0459, -0.0447, -0.0437, ..., 0.0024, 0.0026, 0.0027], + [-0.0457, -0.0449, -0.0437, ..., 0.0023, 0.0025, 0.0024], + [-0.0286, -0.0286, -0.0287, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0017, -0.0012, -0.0013, ..., -0.0017, -0.0018, 0.0003], + [ 0.0017, -0.0016, -0.0016, ..., -0.0017, -0.0016, 0.0008], + [ 0.0070, -0.0070, -0.0069, ..., -0.0038, -0.0037, -0.0009], + ..., + [ 0.0530, -0.0574, -0.0564, ..., -0.0398, -0.0361, -0.0023], + [ 0.0076, -0.0072, -0.0071, ..., -0.0076, -0.0082, -0.0072], + [-0.0232, 0.0238, 0.0236, ..., 0.0216, 0.0214, 0.0126]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.2734e-02, 5.1758e-02, 5.0781e-02, ..., -2.8992e-04, + -2.1935e-04, -3.7193e-04], + [ 4.7607e-02, 4.6143e-02, 4.4678e-02, ..., 4.5967e-04, + 5.6839e-04, 5.9891e-04], + [-4.4434e-02, -4.3213e-02, -4.1992e-02, ..., 1.4210e-04, + 1.9932e-04, 8.9645e-05], + ..., + [ 3.8086e-02, 3.7109e-02, 3.6377e-02, ..., -4.1962e-05, + -1.0967e-04, -8.0109e-05], + [ 5.7861e-02, 5.6641e-02, 5.6152e-02, ..., 3.4714e-04, + 3.8719e-04, 3.7956e-04], + [-4.3701e-02, -4.2969e-02, -4.2236e-02, ..., -1.5259e-04, + -1.9741e-04, -1.5736e-04]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_cross_attn.2.weight': tensor([[ 8.5449e-04, 1.4038e-03, 1.3809e-03, ..., -2.8229e-03, + -3.5706e-03, 7.8678e-05], + [-8.7891e-03, 4.7913e-03, 7.9956e-03, ..., -6.3782e-03, + -5.3101e-03, -7.0190e-04], + [ 1.8234e-03, -2.4986e-04, -8.9645e-04, ..., 4.4632e-04, + 2.8687e-03, -2.7924e-03], + ..., + [ 2.4170e-02, 1.2512e-02, -1.2878e-02, ..., 5.0049e-03, + 1.0132e-02, -1.3046e-03], + [-5.2795e-03, 2.6367e-02, -1.6499e-04, ..., 7.9956e-03, + 3.8818e-02, -2.4292e-02], + [-8.9111e-03, 1.2024e-02, 8.9111e-03, ..., -7.1716e-03, + -3.6469e-03, 1.5991e-02]], dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.1.weight': tensor([[-0.0437, -0.0422, -0.0413, ..., 0.0032, 0.0035, 0.0034], + [-0.0008, 0.0010, 0.0010, ..., 0.0004, 0.0005, 0.0004], + [ 0.0552, 0.0532, 0.0522, ..., -0.0019, -0.0019, -0.0019], + ..., + [ 0.0564, 0.0547, 0.0537, ..., -0.0020, -0.0020, -0.0018], + [ 0.0106, 0.0093, 0.0092, ..., -0.0007, -0.0006, -0.0006], + [-0.0500, -0.0486, -0.0476, ..., 0.0024, 0.0024, 0.0023]], + dtype=torch.bfloat16), 'net.blocks.2.adaln_modulation_mlp.2.weight': tensor([[-0.0273, -0.0060, 0.0178, ..., 0.0111, 0.0077, -0.0182], + [ 0.0034, -0.0017, -0.0023, ..., -0.0014, 0.0016, 0.0017], + [-0.0009, -0.0049, -0.0021, ..., -0.0036, 0.0052, 0.0023], + ..., + [ 0.0415, 0.0010, -0.0156, ..., -0.0097, -0.0019, 0.0161], + [ 0.0003, 0.0090, 0.0043, ..., 0.0073, -0.0089, -0.0048], + [ 0.0042, -0.0170, -0.0103, ..., -0.0128, 0.0173, 0.0098]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_proj.weight': tensor([[-0.0128, 0.0234, 0.0253, ..., -0.0055, 0.0037, -0.0408], + [-0.0029, -0.0024, 0.0052, ..., -0.0020, 0.0045, -0.0206], + [-0.0093, -0.0014, 0.0135, ..., -0.0075, -0.0089, 0.0063], + ..., + [ 0.0060, -0.0057, -0.0042, ..., -0.0045, 0.0106, 0.0017], + [ 0.0012, -0.0145, 0.0063, ..., -0.0327, -0.0079, -0.0267], + [-0.0024, -0.0012, 0.0007, ..., 0.0058, 0.0006, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm.weight': tensor([ 8.1641e-01, 9.8047e-01, 1.0312e+00, 1.0234e+00, 1.0498e-01, + 6.3477e-02, 2.6489e-02, 9.7771e-11, -5.6152e-03, 1.2817e-03, + 1.7578e-02, 8.8867e-02, -4.6997e-03, -6.7444e-03, 3.7193e-04, + 6.1512e-05, 6.2500e-02, 7.3828e-01, 7.9688e-01, -1.4853e-30, + 2.2650e-06, 1.0234e+00, 1.2500e+00, 1.2734e+00, 7.8125e-01, + 7.0312e-01, 9.5703e-01, -1.1230e-02, -7.9349e-07, -1.7889e-17, + -4.0283e-03, -1.0681e-03, 1.9043e-02, 8.4375e-01, 6.5918e-03, + 7.8906e-01, 2.2737e-12, 1.6298e-08, 1.9531e-02, -1.0498e-02, + -7.4463e-03, 3.7486e-08, -4.3365e-09, 1.2578e+00, 9.3359e-01, + 1.0703e+00, 1.0938e+00, 1.1641e+00, 6.2256e-02, 9.5367e-06, + 2.7710e-02, 1.5723e-01, 8.6719e-01, 4.0527e-02, -3.0249e-06, + 9.0332e-02, -3.0060e-03, 9.9609e-02, -6.7230e-09, -8.9722e-03, + -9.8680e-11, 1.8387e-03, -7.9956e-03, 9.0234e-01, 1.0234e+00, + 4.1797e-01, 6.4062e-01, 9.0234e-01, 5.7129e-02, -2.2095e-02, + 4.2419e-03, -2.5153e-12, -1.8433e-02, 1.3123e-02, -1.0300e-03, + 5.0659e-03, 3.2806e-03, 7.1411e-03, -2.9541e-02, 5.2185e-03, + 5.2734e-02, 9.5825e-03, 2.0020e-02, 1.0097e-27, 1.9441e-08, + -5.5847e-03, 3.6914e-01, 2.5195e-01, 1.1016e+00, 1.0391e+00, + 5.1953e-01, -6.5613e-03, -2.2817e-07, 1.2848e-17, -5.2643e-04, + 2.5368e-04, 1.9775e-02, 4.1016e-02, 3.5095e-03, 4.8096e-02, + 1.5916e-10, 1.3562e-08, 9.7168e-02, -1.3649e-05, 1.7944e-02, + -1.3113e-06, 1.3713e-12, 2.2656e-01, 1.0938e+00, 1.0312e+00, + 6.9922e-01, 6.6406e-01, 7.5195e-02, -3.9637e-06, 2.4048e-02, + 8.8281e-01, 1.8677e-02, 8.5547e-01, -2.8312e-06, 6.4468e-04, + 2.1240e-02, 4.3213e-02, 3.9116e-07, -1.6632e-03, -3.4645e-07, + 1.0254e-02, 8.1250e-01, -6.0730e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.k_proj.weight': tensor([[-0.0396, -0.0107, -0.0069, ..., 0.0034, -0.0068, -0.0111], + [-0.0190, -0.0317, -0.0010, ..., 0.0063, -0.0016, -0.0028], + [ 0.0035, 0.0080, 0.0010, ..., -0.0104, 0.0008, -0.0017], + ..., + [ 0.0002, 0.0132, -0.0005, ..., -0.0043, -0.0038, -0.0048], + [ 0.0194, -0.0041, 0.0041, ..., -0.0051, 0.0121, 0.0063], + [ 0.0017, 0.0031, -0.0025, ..., -0.0014, 0.0054, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm.weight': tensor([ 2.4805e-01, 9.5312e-01, 1.0547e+00, 9.9219e-01, 1.0889e-01, + 7.0801e-02, 2.2339e-02, -8.0036e-11, 3.8757e-03, -1.9531e-03, + 1.7334e-02, 8.9844e-02, -8.7891e-03, 3.1738e-02, -6.0558e-05, + 6.0081e-05, 7.5684e-03, 7.3828e-01, 7.9297e-01, 2.0831e-30, + 9.9838e-07, 1.0234e+00, 1.1250e+00, 1.1562e+00, 3.5858e-03, + 1.4282e-02, 9.8047e-01, -8.9722e-03, -5.1409e-07, 3.0791e-17, + -4.1504e-03, 1.1215e-03, -1.6113e-02, 8.3984e-01, 3.9673e-03, + 7.8516e-01, 1.2150e-12, 1.6997e-08, -2.1484e-02, -1.0620e-02, + 7.5073e-03, -4.1211e-08, 3.8017e-10, 1.1016e+00, 2.9883e-01, + 1.0234e+00, 1.1562e+00, 1.1797e+00, -7.0572e-04, 2.7657e-05, + 1.0925e-02, 1.0529e-03, 8.6719e-01, 5.7373e-02, -5.5507e-07, + 8.9844e-02, -2.5787e-03, 1.0010e-01, 2.2119e-09, 7.5195e-02, + 1.9554e-10, 1.3367e-02, -2.3804e-02, 9.0234e-01, 1.0547e+00, + 7.4219e-01, 4.3555e-01, 9.1016e-01, 3.5889e-02, -1.1597e-02, + -2.0752e-02, 3.1974e-12, -1.8066e-02, 1.3367e-02, 1.4282e-02, + 4.0039e-02, 4.0283e-03, 2.8198e-02, -4.1992e-02, -5.1575e-03, + -1.3123e-02, 1.0315e-02, 1.1597e-02, -9.9081e-28, 4.9360e-08, + 6.9885e-03, 1.0000e+00, 9.6484e-01, 1.1875e+00, 1.0938e+00, + 2.5757e-02, 1.2024e-02, 5.6624e-07, -1.1330e-17, 6.2561e-04, + 2.4986e-04, 2.9602e-03, -1.2436e-03, -3.2959e-03, 3.9368e-03, + 1.5189e-10, -1.2806e-08, 9.6680e-02, -1.4782e-05, 1.7944e-02, + 9.6858e-07, -4.4338e-12, 9.5703e-01, 1.2266e+00, 1.0547e+00, + 2.2217e-02, 6.2109e-01, 7.2937e-03, 1.4082e-06, -8.1177e-03, + 8.8281e-01, 3.6316e-03, 8.5938e-01, -2.6375e-06, 6.7711e-05, + 2.1240e-02, 4.2725e-02, 3.8743e-07, 4.1504e-02, 1.9837e-07, + 2.1240e-02, 8.1250e-01, -1.4343e-03], dtype=torch.bfloat16), 'net.blocks.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.self_attn.v_proj.weight': tensor([[ 1.4954e-03, -1.3367e-02, 5.6763e-03, ..., 4.5654e-02, + -7.9956e-03, 5.5542e-03], + [ 5.8289e-03, 1.7456e-02, -8.0566e-03, ..., -2.1729e-02, + -4.1199e-03, -8.0566e-03], + [-3.9816e-05, 1.6479e-02, -1.2573e-02, ..., -5.4321e-03, + 3.5706e-03, 9.2773e-03], + ..., + [-1.4496e-04, 5.4626e-03, -6.9580e-03, ..., 9.2773e-03, + 3.8757e-03, 2.3746e-04], + [ 2.3804e-03, 1.8066e-02, -4.5471e-03, ..., -2.1210e-03, + 3.4943e-03, 3.9978e-03], + [-6.3171e-03, -9.5215e-03, 6.8359e-03, ..., 2.1057e-03, + -2.6245e-03, -1.5747e-02]], dtype=torch.bfloat16), 'net.blocks.3.self_attn.output_proj.weight': tensor([[ 0.0192, 0.0210, -0.0076, ..., 0.0072, 0.0312, -0.0110], + [ 0.0141, 0.0040, 0.0036, ..., -0.0184, -0.0089, -0.0036], + [-0.0095, 0.0167, -0.0125, ..., -0.0013, -0.0047, 0.0043], + ..., + [-0.0294, -0.0309, -0.0076, ..., 0.0094, -0.0043, 0.0026], + [-0.0092, 0.0095, -0.0223, ..., -0.0072, 0.0283, 0.0001], + [-0.0378, 0.0050, 0.0172, ..., 0.0244, -0.0037, -0.0162]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_proj.weight': tensor([[ 0.0017, -0.0059, -0.0021, ..., -0.0069, -0.0047, 0.0129], + [ 0.0036, -0.0092, 0.0008, ..., -0.0070, 0.0013, 0.0107], + [ 0.0030, 0.0041, -0.0029, ..., 0.0099, 0.0176, -0.0055], + ..., + [ 0.0096, -0.0032, -0.0016, ..., -0.0008, -0.0021, 0.0040], + [ 0.0060, 0.0025, -0.0006, ..., -0.0023, 0.0008, 0.0047], + [-0.0068, -0.0025, -0.0043, ..., -0.0052, 0.0055, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0376, 0.1533, 0.1328, 0.4727, + 0.0366, 0.0654, 0.0187, 0.0461, 0.0688, 0.0557, 0.1943, 0.4199, 0.0293, + 0.7188, 0.0425, 0.0396, 0.1602, 0.0205, 0.0889, 0.0525, 0.3828, 0.4473, + 0.2578, 0.0757, 0.0415, 0.0201, 0.2139, 0.6562, 0.2500, 0.0430, 0.0325, + 0.0186, 0.0986, 0.0253, 0.0376, 0.0459, 0.1245, 0.0234, 0.0403, 0.2100, + 0.0391, 0.3457, 0.0240, 0.0549, 0.0217, 0.0198, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0405, 0.1436, 0.0256, 0.0918, 0.4277, 0.0159, + 0.4688, 0.1699, 0.3730, 0.0918, 0.0547, 0.0718, 0.4180, 0.0283, 0.4395, + 0.0195, 0.0391, 0.1426, 0.0923, 0.0161, 0.0967, 0.6133, 0.0417, 0.0923, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0737, 0.0615, 0.0986, 0.0205, 0.1553, + 0.0325, 0.1729, 0.0471, 0.0693, 0.0427, 0.1147, 0.0104, 0.3301, 0.4609, + 0.0476, 0.0645, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0315, 0.5078, + 0.4844, 0.0374, 0.0128, 0.6875, 0.0272, 0.0156, 0.4258, 0.2109, 0.2793, + 0.6836, 0.0459, 0.0874, 0.0757, 0.4961, 0.5469, 0.0154, 0.0261, 0.1934, + 0.0334, 0.0349], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_proj.weight': tensor([[-0.0020, 0.0020, 0.0036, ..., 0.0004, 0.0142, -0.0033], + [ 0.0087, -0.0002, -0.0033, ..., -0.0048, -0.0025, 0.0048], + [-0.0145, 0.0078, -0.0141, ..., 0.0018, 0.0082, 0.0035], + ..., + [ 0.0011, 0.0030, 0.0008, ..., -0.0021, -0.0049, 0.0034], + [-0.0150, -0.0020, -0.0144, ..., 0.0085, -0.0055, 0.0042], + [ 0.0194, -0.0016, 0.0073, ..., -0.0100, 0.0031, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm.weight': tensor([0.1060, 0.0238, 0.3848, 0.0613, 0.1650, 0.0378, 0.1533, 0.1328, 0.4727, + 0.0371, 0.0664, 0.0187, 0.0461, 0.0693, 0.0559, 0.1963, 0.4199, 0.0294, + 0.7148, 0.0422, 0.0393, 0.1602, 0.0206, 0.0884, 0.0520, 0.3789, 0.4473, + 0.2559, 0.0757, 0.0413, 0.0211, 0.2158, 0.6562, 0.2500, 0.0437, 0.0325, + 0.0183, 0.0986, 0.0254, 0.0371, 0.0454, 0.1260, 0.0237, 0.0408, 0.2090, + 0.0391, 0.3457, 0.0242, 0.0549, 0.0211, 0.0203, 0.3770, 0.4453, 0.3359, + 0.0962, 0.5195, 0.5703, 0.0403, 0.1436, 0.0259, 0.0918, 0.4277, 0.0161, + 0.4688, 0.1689, 0.3730, 0.0918, 0.0547, 0.0718, 0.4199, 0.0282, 0.4395, + 0.0193, 0.0391, 0.1426, 0.0913, 0.0161, 0.0977, 0.6133, 0.0413, 0.0913, + 0.0918, 0.0349, 0.0522, 0.3887, 0.0742, 0.0603, 0.0981, 0.0209, 0.1572, + 0.0337, 0.1719, 0.0471, 0.0698, 0.0430, 0.1157, 0.0103, 0.3301, 0.4609, + 0.0474, 0.0649, 0.3027, 0.0344, 0.1270, 0.0347, 0.2275, 0.0312, 0.5078, + 0.4844, 0.0369, 0.0128, 0.6875, 0.0267, 0.0159, 0.4277, 0.2109, 0.2793, + 0.6797, 0.0454, 0.0854, 0.0757, 0.4961, 0.5469, 0.0151, 0.0261, 0.1934, + 0.0344, 0.0354], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.v_proj.weight': tensor([[ 0.0092, 0.0027, -0.0237, ..., 0.0068, 0.0009, -0.0115], + [-0.0079, 0.0085, -0.0004, ..., 0.0020, -0.0011, -0.0236], + [ 0.0225, -0.0062, -0.0073, ..., -0.0049, -0.0069, -0.0386], + ..., + [-0.0005, -0.0003, 0.0184, ..., -0.0110, -0.0229, 0.0050], + [-0.0327, 0.0021, -0.0215, ..., -0.0035, 0.0065, -0.0164], + [-0.0084, 0.0024, 0.0061, ..., -0.0078, 0.0115, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0013, -0.0094, ..., -0.0084, 0.0010, 0.0067], + [-0.0238, -0.0147, -0.0074, ..., 0.0026, -0.0160, -0.0192], + [ 0.0014, 0.0038, 0.0025, ..., -0.0005, 0.0044, -0.0022], + ..., + [ 0.0145, -0.0145, 0.0261, ..., 0.0184, -0.0092, -0.0053], + [-0.0141, -0.0135, 0.0114, ..., 0.0183, -0.0149, 0.0251], + [-0.0094, -0.0153, 0.0273, ..., -0.0097, 0.0162, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img.weight': tensor([[ 0.0062, -0.0061, 0.0282, ..., -0.0111, -0.0038, -0.0182], + [ 0.0022, 0.0126, -0.0315, ..., -0.0457, -0.0079, 0.0105], + [-0.0009, 0.0261, 0.0245, ..., 0.0063, 0.0074, 0.0461], + ..., + [ 0.0170, -0.0254, -0.0439, ..., 0.0156, 0.0034, -0.0067], + [ 0.0076, -0.0371, 0.0356, ..., -0.0082, 0.0410, -0.0178], + [ 0.0178, -0.0146, 0.0562, ..., -0.0291, -0.0267, 0.0298]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.v_img.weight': tensor([[ 0.0078, 0.0208, 0.0276, ..., 0.0136, 0.0151, 0.0083], + [-0.0062, 0.0025, -0.0049, ..., 0.0145, 0.0005, 0.0048], + [ 0.0056, 0.0021, 0.0242, ..., -0.0120, -0.0070, -0.0248], + ..., + [ 0.0245, 0.0211, -0.0059, ..., 0.0248, -0.0095, 0.0381], + [-0.0308, 0.0173, -0.0310, ..., -0.0010, -0.0295, 0.0400], + [-0.0114, 0.0130, 0.0305, ..., -0.0107, 0.0366, 0.0228]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img.weight': tensor([[-0.0049, 0.0381, -0.0239, ..., 0.0211, 0.0483, 0.0339], + [-0.0146, -0.0043, 0.0288, ..., 0.0150, 0.0583, 0.0001], + [-0.0281, 0.0156, -0.0386, ..., -0.0444, -0.0093, -0.0013], + ..., + [ 0.0030, 0.0157, 0.0217, ..., -0.0005, 0.0115, 0.0151], + [ 0.0081, 0.0195, -0.0283, ..., -0.0009, 0.0021, -0.0342], + [-0.0162, -0.0195, -0.0013, ..., -0.0024, 0.0165, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.3.mlp.layer1.weight': tensor([[ 0.0198, 0.0081, -0.0510, ..., 0.0051, 0.0066, 0.0024], + [-0.0070, 0.0126, 0.0026, ..., -0.0121, -0.0118, -0.0071], + [ 0.0040, 0.0256, -0.0208, ..., 0.0282, -0.0098, 0.0159], + ..., + [-0.0027, 0.0096, 0.0762, ..., -0.0067, 0.0129, -0.0094], + [ 0.0052, -0.0089, -0.0115, ..., -0.0124, -0.0012, -0.0078], + [ 0.0057, 0.0079, 0.0020, ..., -0.0206, -0.0190, -0.0126]], + dtype=torch.bfloat16), 'net.blocks.3.mlp.layer2.weight': tensor([[ 0.0082, -0.0020, 0.0018, ..., -0.0140, -0.0083, -0.0025], + [ 0.0036, 0.0118, 0.0045, ..., 0.0023, 0.0136, 0.0107], + [-0.0004, -0.0035, -0.0001, ..., -0.0054, 0.0090, -0.0013], + ..., + [ 0.0019, 0.0088, 0.0104, ..., -0.0063, -0.0020, 0.0023], + [ 0.0093, 0.0004, -0.0005, ..., 0.0090, 0.0089, 0.0042], + [-0.0099, 0.0014, 0.0063, ..., 0.0093, 0.0126, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0542, -0.0532, -0.0522, ..., 0.0012, 0.0010, 0.0011], + [-0.0474, -0.0464, -0.0452, ..., 0.0029, 0.0030, 0.0030], + [-0.0469, -0.0457, -0.0449, ..., 0.0024, 0.0027, 0.0026], + ..., + [ 0.0184, 0.0176, 0.0166, ..., -0.0036, -0.0037, -0.0036], + [ 0.0038, 0.0028, 0.0023, ..., -0.0031, -0.0030, -0.0030], + [-0.0469, -0.0459, -0.0452, ..., 0.0025, 0.0027, 0.0028]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, 0.0042, 0.0037, ..., 0.0002, 0.0052, 0.0043], + [-0.0011, -0.0028, 0.0004, ..., 0.0015, 0.0016, -0.0012], + [-0.0035, -0.0054, -0.0044, ..., -0.0011, -0.0003, -0.0052], + ..., + [-0.0270, -0.0732, -0.0603, ..., 0.0227, 0.0243, -0.0679], + [-0.0104, -0.0090, -0.0092, ..., -0.0021, -0.0063, -0.0084], + [ 0.0095, 0.0055, 0.0069, ..., 0.0089, 0.0117, 0.0063]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.1.weight': tensor([[-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., 3.6240e-05, + 2.0742e-05, 3.0518e-05], + [ 2.2217e-02, 2.1729e-02, 2.1362e-02, ..., -1.4210e-04, + -1.1492e-04, 4.9174e-06], + [ 1.4832e-02, 1.4160e-02, 1.3672e-02, ..., 1.3733e-04, + 1.6308e-04, 6.7711e-05], + ..., + [ 5.8105e-02, 5.6641e-02, 5.5664e-02, ..., -3.5286e-04, + -3.4714e-04, -2.8038e-04], + [ 6.0547e-02, 5.9326e-02, 5.7617e-02, ..., -3.4714e-04, + -2.8801e-04, -3.3379e-04], + [-1.5076e-02, -1.4709e-02, -1.4282e-02, ..., 9.2506e-05, + 2.4414e-04, 2.6512e-04]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0066, -0.0026, -0.0058, ..., 0.0339, 0.0376, 0.0074], + [ 0.0003, -0.0026, 0.0042, ..., 0.0004, 0.0003, 0.0067], + [-0.0029, 0.0007, -0.0010, ..., -0.0045, -0.0052, -0.0023], + ..., + [-0.0159, 0.0039, 0.0309, ..., 0.0063, 0.0165, 0.0031], + [-0.0330, 0.0228, 0.0238, ..., -0.0089, -0.0087, 0.0052], + [ 0.0188, -0.0135, -0.0016, ..., 0.0071, 0.0112, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.1.weight': tensor([[-6.2256e-02, -5.9814e-02, -5.9082e-02, ..., 7.1526e-05, + 1.9455e-04, 1.4114e-04], + [ 9.4238e-02, 9.1309e-02, 8.8867e-02, ..., 8.8215e-05, + 1.6499e-04, 1.5831e-04], + [-1.2207e-01, -1.1865e-01, -1.1670e-01, ..., -4.4250e-04, + -3.6621e-04, -4.0436e-04], + ..., + [-6.5918e-02, -6.5430e-02, -6.4453e-02, ..., 1.0490e-04, + 2.0981e-04, 1.0586e-04], + [-6.1279e-02, -6.0059e-02, -5.8838e-02, ..., 1.4496e-04, + 8.5831e-06, 1.7524e-05], + [ 6.8848e-02, 6.7871e-02, 6.6895e-02, ..., -7.9155e-05, + -1.1110e-04, -7.2479e-05]], dtype=torch.bfloat16), 'net.blocks.3.adaln_modulation_mlp.2.weight': tensor([[-7.2327e-03, -3.1738e-03, -2.6550e-03, ..., -9.1553e-03, + -1.2878e-02, 1.0681e-02], + [-1.0605e-03, -1.9226e-03, 1.1673e-03, ..., -1.1139e-03, + -6.4850e-05, 5.1498e-04], + [-2.4872e-03, 1.1368e-03, -6.2866e-03, ..., -3.4332e-03, + -3.0670e-03, 3.0060e-03], + ..., + [-5.6152e-02, 1.0498e-02, -7.5195e-02, ..., -4.3945e-02, + -2.2583e-02, 3.5889e-02], + [-2.4719e-03, -1.6113e-02, 2.3193e-02, ..., -3.0670e-03, + -3.5400e-03, 3.4485e-03], + [ 1.1353e-02, 1.2054e-03, 1.3855e-02, ..., 9.8267e-03, + 4.8828e-03, -7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_proj.weight': tensor([[ 0.0679, -0.0041, -0.0007, ..., -0.0242, -0.0014, 0.0693], + [ 0.0024, -0.0037, -0.0018, ..., 0.0078, -0.0031, -0.0030], + [-0.0112, 0.0114, -0.0004, ..., -0.0222, 0.0006, -0.0146], + ..., + [-0.0028, 0.0146, 0.0006, ..., 0.0070, -0.0154, 0.0126], + [-0.0014, 0.0045, -0.0017, ..., 0.0006, -0.0025, 0.0007], + [-0.0032, 0.0020, -0.0031, ..., -0.0014, 0.0005, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm.weight': tensor([ 1.5781e+00, 2.3828e-01, 1.2656e+00, 1.2891e+00, 1.2031e+00, + 2.0874e-02, 5.6152e-02, 5.4443e-02, 7.5000e-01, 8.7500e-01, + 6.4453e-01, 3.6621e-02, 9.8438e-01, 1.9989e-03, 1.6699e-01, + 8.7109e-01, 5.4321e-03, 2.5024e-02, 8.9062e-01, 2.2827e-02, + 7.0312e-01, 2.0630e-02, 3.3594e-01, 7.0703e-01, 1.1562e+00, + 1.2734e+00, 1.2891e+00, 2.5977e-01, 1.0547e-01, 4.0283e-02, + 9.0234e-01, 4.2236e-02, 4.1748e-02, 7.9688e-01, 1.0078e+00, + 9.8633e-02, 1.5918e-01, -2.2583e-03, 7.3047e-01, 2.1606e-02, + 8.7891e-01, 4.9072e-02, 6.3281e-01, 1.5938e+00, 1.5391e+00, + 8.8672e-01, 1.4609e+00, 4.3750e-01, 1.3306e-02, 5.6152e-02, + 6.0303e-02, 6.7749e-03, 4.8096e-02, 4.8340e-02, 1.0938e-01, + 5.3955e-02, 1.0449e-01, 5.6396e-02, 5.3711e-02, 8.2520e-02, + 1.0840e-01, 9.0625e-01, 1.7944e-02, 8.3594e-01, 4.6875e-01, + 1.4609e+00, 1.2422e+00, 1.1172e+00, 4.0430e-01, 1.0889e-01, + 1.2891e-01, 9.2188e-01, 4.8340e-02, 2.1973e-01, 3.8086e-01, + 1.9434e-01, 2.3926e-01, 1.0000e+00, 3.2227e-02, 9.8438e-01, + 7.1484e-01, -7.6771e-05, 4.4922e-02, 4.0820e-01, 2.8198e-02, + 1.9653e-02, 1.5781e+00, 1.4062e+00, 1.3906e+00, 5.1953e-01, + 3.1250e-01, 1.9141e-01, 2.8516e-01, 5.4199e-02, 2.0020e-01, + 9.2188e-01, 1.1230e-01, 8.7500e-01, 6.0547e-02, 6.1328e-01, + 2.2266e-01, 9.1309e-02, 3.0273e-02, 9.1406e-01, 2.8564e-02, + 1.0312e+00, 7.5391e-01, 4.4531e-01, 1.0449e-01, 1.3281e+00, + 1.3203e+00, 1.4531e+00, 2.9688e-01, 3.1641e-01, 1.5381e-02, + 6.5918e-02, 6.3965e-02, 4.2419e-03, 8.3203e-01, -5.4932e-04, + 3.2959e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1118e-02, + 7.3828e-01, 1.8555e-02, -2.3346e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.k_proj.weight': tensor([[ 0.1118, -0.0287, -0.0009, ..., -0.0262, -0.0160, 0.0825], + [ 0.0092, 0.0043, 0.0009, ..., 0.0010, 0.0011, -0.0053], + [-0.0104, 0.0194, -0.0020, ..., -0.0215, -0.0131, -0.0198], + ..., + [-0.0023, 0.0078, 0.0003, ..., 0.0039, -0.0297, -0.0219], + [ 0.0036, 0.0058, 0.0039, ..., -0.0002, 0.0085, 0.0020], + [-0.0020, 0.0043, 0.0042, ..., -0.0026, 0.0107, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm.weight': tensor([ 1.3438e+00, 1.0156e+00, 1.2656e+00, 1.2344e+00, 1.2109e+00, + 3.1738e-02, 5.5420e-02, 1.3306e-02, 7.5000e-01, 8.7500e-01, + 6.4844e-01, 4.1016e-02, 9.8438e-01, 2.3682e-02, 1.6699e-01, + 8.7109e-01, 2.4292e-02, 2.4414e-02, 8.9453e-01, 2.2949e-02, + 6.9922e-01, 2.0142e-02, 8.9453e-01, 5.9766e-01, 1.2734e+00, + 1.2891e+00, 1.1953e+00, 2.6172e-01, 1.0107e-01, 4.2725e-02, + 9.0625e-01, 4.1016e-02, 4.7852e-02, 7.9297e-01, 1.0078e+00, + 1.0645e-01, 1.5820e-01, -1.0529e-03, 7.3047e-01, 2.2217e-02, + 8.7891e-01, 4.9561e-02, 6.3281e-01, 1.3672e+00, 1.3047e+00, + 3.3203e-01, 1.3203e+00, 6.2891e-01, 7.6172e-02, 1.9653e-02, + 5.7129e-02, 7.7515e-03, 4.8096e-02, 4.8828e-02, 6.4453e-02, + 5.4199e-02, 1.0498e-01, 5.3955e-02, 6.0547e-02, 7.8125e-02, + 1.0742e-01, 9.0625e-01, 1.3977e-02, 8.3594e-01, 8.7891e-01, + 1.3047e+00, 1.1953e+00, 1.0703e+00, 2.0996e-01, 1.0742e-01, + 1.3086e-01, 9.2188e-01, 2.0142e-02, 1.8652e-01, 3.7891e-01, + 1.9336e-01, 2.4219e-01, 1.0000e+00, 2.7466e-02, 9.8438e-01, + 7.1875e-01, 3.1948e-05, 4.1748e-02, 4.0625e-01, 2.8809e-02, + 2.0264e-02, 1.4375e+00, 1.4609e+00, 1.2422e+00, 3.6914e-01, + 4.7070e-01, 2.0605e-01, 2.9102e-01, 5.2979e-02, 1.0156e-01, + 9.2188e-01, 1.2158e-01, 8.7109e-01, 6.0547e-02, 6.1328e-01, + 2.2168e-01, 9.0820e-02, 2.9297e-02, 9.1406e-01, 2.9785e-02, + 1.0312e+00, 7.5391e-01, 1.0391e+00, 1.0938e+00, 1.4141e+00, + 1.2969e+00, 1.3984e+00, 3.1250e-01, 3.1641e-01, 2.7832e-02, + 6.5918e-02, 6.6895e-02, 4.2725e-03, 8.2812e-01, -4.2915e-05, + 2.9297e-02, 1.5527e-01, 8.1641e-01, 7.3828e-01, 2.1240e-02, + 7.3828e-01, 1.2451e-02, 2.1973e-03], dtype=torch.bfloat16), 'net.blocks.4.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.self_attn.v_proj.weight': tensor([[ 0.0032, -0.0149, 0.0535, ..., 0.0107, 0.0082, -0.0003], + [-0.0023, -0.0184, -0.0179, ..., -0.0027, 0.0082, 0.0045], + [ 0.0009, -0.0040, 0.0184, ..., -0.0067, 0.0049, -0.0025], + ..., + [ 0.0059, 0.0093, -0.0192, ..., -0.0081, 0.0184, -0.0103], + [-0.0012, 0.0021, 0.0101, ..., -0.0072, -0.0052, 0.0018], + [-0.0049, 0.0052, -0.0006, ..., 0.0117, -0.0154, 0.0093]], + dtype=torch.bfloat16), 'net.blocks.4.self_attn.output_proj.weight': tensor([[ 0.0194, 0.0041, 0.0238, ..., 0.0147, -0.0037, -0.0060], + [ 0.0005, -0.0013, -0.0037, ..., -0.0026, -0.0166, 0.0118], + [ 0.0119, -0.0143, -0.0195, ..., 0.0079, 0.0041, -0.0131], + ..., + [-0.0042, -0.0091, -0.0135, ..., -0.0201, 0.0056, -0.0050], + [-0.0062, 0.0139, 0.0085, ..., 0.0090, -0.0106, 0.0085], + [ 0.0072, 0.0002, -0.0135, ..., 0.0079, -0.0044, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_proj.weight': tensor([[-0.0003, 0.0112, 0.0032, ..., 0.0159, 0.0157, 0.0103], + [ 0.0103, 0.0055, -0.0017, ..., -0.0021, 0.0010, -0.0034], + [-0.0081, 0.0047, -0.0061, ..., 0.0024, -0.0109, 0.0070], + ..., + [ 0.0060, 0.0054, 0.0013, ..., 0.0078, -0.0121, 0.0122], + [ 0.0004, 0.0035, -0.0023, ..., -0.0046, -0.0015, 0.0060], + [ 0.0048, 0.0067, 0.0050, ..., -0.0079, 0.0060, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm.weight': tensor([0.5938, 0.0129, 0.4688, 0.0156, 0.0972, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0262, 0.0085, 0.4609, 0.0284, 0.3691, + 0.0023, 0.0245, 0.0437, 0.0767, 0.3301, 0.0054, 0.2852, 0.0092, 0.0120, + 0.0762, 0.0703, 0.5664, 0.1299, 0.5820, 0.2178, 0.1465, 0.1177, 0.0388, + 0.0228, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0342, 0.1040, + 0.0168, 0.0143, 0.1641, 0.0256, 0.2578, 0.5625, 0.0255, 0.0151, 0.0432, + 0.2051, 0.1445, 0.0026, 0.0064, 0.1719, 0.1309, 0.2031, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0515, 0.0562, 0.1787, 0.0342, 0.2656, + 0.2305, 0.0089, 0.2031, 0.0188, 0.1553, 0.0275, 0.5781, 0.0184, 0.0141, + 0.6914, 0.0146, 0.0045, 0.0032, 0.0820, 0.0047, 0.0354, 0.1338, 0.0957, + 0.4414, 0.0137, 0.0258, 0.4004, 0.0090, 0.0278, 0.5898, 0.0277, 0.5469, + 0.4824, 0.4219, 0.0154, 0.0452, 0.5508, 0.1133, 0.1338, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0525, 0.6094, 0.0232, 0.6211, 0.0161, 0.0199, 0.0386, + 0.0030, 0.0115, 0.0239, 0.0771, 0.5195, 0.0708, 0.6289, 0.0415, 0.3086, + 0.0294, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_proj.weight': tensor([[-0.0030, 0.0004, -0.0074, ..., -0.0035, 0.0067, -0.0150], + [-0.0073, -0.0066, -0.0025, ..., 0.0120, 0.0007, -0.0134], + [-0.0176, 0.0092, 0.0019, ..., -0.0026, -0.0033, 0.0060], + ..., + [ 0.0038, 0.0049, -0.0008, ..., 0.0075, 0.0013, 0.0047], + [ 0.0014, -0.0012, -0.0003, ..., -0.0061, -0.0110, 0.0046], + [ 0.0129, 0.0017, 0.0074, ..., -0.0071, -0.0017, -0.0015]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm.weight': tensor([0.6016, 0.0131, 0.4688, 0.0155, 0.0977, 0.1289, 0.0116, 0.6055, 0.0542, + 0.4727, 0.5391, 0.6094, 0.4297, 0.0266, 0.0086, 0.4609, 0.0286, 0.3691, + 0.0024, 0.0248, 0.0439, 0.0757, 0.3301, 0.0056, 0.2852, 0.0091, 0.0120, + 0.0767, 0.0708, 0.5625, 0.1309, 0.5820, 0.2178, 0.1475, 0.1177, 0.0386, + 0.0229, 0.1235, 0.0806, 0.6016, 0.0306, 0.5664, 0.6133, 0.0339, 0.1050, + 0.0170, 0.0144, 0.1631, 0.0254, 0.2578, 0.5625, 0.0254, 0.0150, 0.0432, + 0.2041, 0.1465, 0.0025, 0.0066, 0.1709, 0.1309, 0.2051, 0.3301, 0.0786, + 0.0083, 0.1943, 0.1592, 0.0908, 0.0505, 0.0559, 0.1787, 0.0344, 0.2656, + 0.2324, 0.0090, 0.2031, 0.0184, 0.1533, 0.0275, 0.5781, 0.0184, 0.0142, + 0.6914, 0.0143, 0.0045, 0.0032, 0.0820, 0.0047, 0.0352, 0.1348, 0.0957, + 0.4414, 0.0135, 0.0261, 0.4004, 0.0088, 0.0277, 0.5938, 0.0280, 0.5469, + 0.4824, 0.4219, 0.0153, 0.0454, 0.5508, 0.1128, 0.1348, 0.6797, 0.1689, + 0.7109, 0.1826, 0.0522, 0.6094, 0.0237, 0.6211, 0.0162, 0.0194, 0.0386, + 0.0029, 0.0114, 0.0243, 0.0771, 0.5195, 0.0713, 0.6289, 0.0415, 0.3086, + 0.0289, 0.1260], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.v_proj.weight': tensor([[ 0.0459, 0.0110, 0.0134, ..., 0.0081, -0.0117, -0.0175], + [-0.0050, 0.0030, 0.0121, ..., 0.0038, 0.0058, -0.0133], + [-0.0042, 0.0018, 0.0145, ..., 0.0050, 0.0121, -0.0068], + ..., + [-0.0190, -0.0017, 0.0050, ..., 0.0064, 0.0186, -0.0018], + [-0.0058, 0.0002, 0.0154, ..., -0.0029, -0.0057, 0.0092], + [ 0.0201, -0.0057, -0.0034, ..., -0.0019, -0.0129, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.output_proj.weight': tensor([[ 0.0092, 0.0023, -0.0025, ..., 0.0283, 0.0081, 0.0120], + [-0.0082, -0.0092, 0.0030, ..., 0.0126, -0.0146, -0.0173], + [ 0.0085, -0.0129, -0.0097, ..., 0.0041, 0.0172, -0.0171], + ..., + [ 0.0014, -0.0023, -0.0147, ..., 0.0076, 0.0125, -0.0078], + [-0.0099, 0.0146, 0.0148, ..., 0.0149, 0.0137, -0.0168], + [ 0.0112, 0.0089, -0.0121, ..., -0.0021, -0.0076, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img.weight': tensor([[ 0.0146, -0.0138, 0.0459, ..., -0.0018, -0.0320, 0.0278], + [ 0.0019, 0.0045, 0.0270, ..., -0.0220, -0.0183, -0.0271], + [ 0.0221, 0.0217, -0.0201, ..., 0.0030, 0.0062, -0.0053], + ..., + [-0.0258, 0.0034, 0.0215, ..., -0.0018, 0.0237, -0.0266], + [-0.0200, 0.0166, 0.0349, ..., 0.0240, -0.0266, -0.0073], + [ 0.0026, -0.0082, -0.0130, ..., -0.0334, 0.0057, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.v_img.weight': tensor([[-0.0435, -0.0155, -0.0454, ..., -0.0134, -0.0128, 0.0400], + [ 0.0291, -0.0189, -0.0110, ..., -0.0288, 0.0025, -0.0186], + [ 0.0043, 0.0547, -0.0019, ..., 0.0120, 0.0400, -0.0215], + ..., + [ 0.0275, 0.0042, -0.0269, ..., -0.0096, -0.0287, 0.0150], + [ 0.0101, 0.0415, -0.0200, ..., -0.0483, 0.0289, -0.0254], + [-0.0439, 0.0095, 0.0108, ..., -0.0058, 0.0703, -0.0052]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img.weight': tensor([[ 0.0203, 0.0091, -0.0315, ..., 0.0074, 0.0254, 0.0359], + [ 0.0079, 0.0051, 0.0040, ..., -0.0119, -0.0233, -0.0065], + [ 0.0046, -0.0037, -0.0118, ..., -0.0125, -0.0072, 0.0084], + ..., + [-0.0354, 0.0305, -0.0596, ..., -0.0334, 0.0400, -0.0026], + [ 0.0339, -0.0066, 0.0223, ..., 0.0410, 0.0002, 0.0128], + [-0.0049, -0.0439, -0.0486, ..., 0.0137, -0.0129, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.4.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.4.mlp.layer1.weight': tensor([[ 0.0042, 0.0126, -0.0310, ..., 0.0141, 0.0121, -0.0007], + [-0.0023, -0.0378, 0.0059, ..., 0.0118, 0.0190, -0.0095], + [-0.0035, -0.0106, 0.0060, ..., 0.0062, 0.0029, 0.0024], + ..., + [ 0.0061, 0.0172, 0.0089, ..., -0.0334, -0.0148, 0.0171], + [-0.0221, -0.0271, 0.0049, ..., 0.0131, 0.0132, -0.0021], + [ 0.0090, -0.0297, 0.0016, ..., 0.0030, 0.0098, -0.0022]], + dtype=torch.bfloat16), 'net.blocks.4.mlp.layer2.weight': tensor([[ 0.0006, -0.0160, 0.0238, ..., -0.0066, 0.0168, 0.0209], + [ 0.0052, -0.0231, 0.0133, ..., -0.0101, 0.0089, -0.0039], + [ 0.0061, -0.0159, 0.0014, ..., -0.0189, -0.0015, -0.0003], + ..., + [ 0.0002, 0.0188, -0.0160, ..., 0.0107, 0.0028, 0.0026], + [ 0.0078, 0.0334, -0.0058, ..., 0.0006, -0.0143, 0.0184], + [ 0.0058, -0.0055, -0.0019, ..., 0.0269, -0.0178, 0.0322]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0303e-02, -5.8838e-02, ..., -5.0068e-06, + -1.3828e-05, -6.9141e-05], + [-6.0303e-02, -5.8838e-02, -5.6641e-02, ..., -1.2970e-04, + -1.1587e-04, -1.4496e-04], + [ 6.1279e-02, 5.9570e-02, 5.8350e-02, ..., 1.6689e-05, + -5.1498e-05, 1.1396e-04], + ..., + [ 5.7617e-02, 5.6396e-02, 5.5176e-02, ..., -5.2691e-05, + 2.3365e-05, 6.8665e-05], + [ 1.3855e-02, 1.3000e-02, 1.2207e-02, ..., -1.7014e-03, + -1.7014e-03, -1.7242e-03], + [-5.9570e-02, -5.7861e-02, -5.6885e-02, ..., 1.5974e-05, + -6.0558e-05, 1.5378e-05]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_self_attn.2.weight': tensor([[ 1.9287e-02, 1.5564e-02, -1.6479e-02, ..., -2.5146e-02, + -7.5989e-03, 2.2827e-02], + [ 1.7242e-03, 1.3962e-03, -1.0452e-03, ..., -2.5024e-03, + 8.6975e-04, 2.4872e-03], + [-6.9618e-05, -4.0054e-05, 6.2943e-05, ..., 1.7762e-05, + -9.5367e-04, -5.7936e-05], + ..., + [ 2.8198e-02, 2.7222e-02, -2.7344e-02, ..., -2.8809e-02, + 1.3916e-02, 2.8564e-02], + [-2.5177e-03, -4.0894e-03, 4.4556e-03, ..., 3.7842e-03, + -2.6093e-03, -4.3030e-03], + [ 5.7678e-03, 4.5471e-03, -5.4932e-03, ..., -7.9956e-03, + -4.4861e-03, 6.8970e-03]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.1.weight': tensor([[ 1.4526e-02, 1.4465e-02, 1.4221e-02, ..., -1.7643e-04, + -6.6280e-05, -2.1648e-04], + [-2.0020e-02, -1.9775e-02, -1.9165e-02, ..., 1.2302e-04, + 1.6022e-04, 1.7166e-04], + [-2.4170e-02, -2.3804e-02, -2.3438e-02, ..., 1.8501e-04, + 6.7234e-05, 1.4210e-04], + ..., + [-1.2894e-03, -1.2131e-03, -1.0071e-03, ..., 1.8311e-04, + 1.5259e-04, 1.1635e-04], + [ 1.4648e-02, 1.4160e-02, 1.3672e-02, ..., -1.9550e-04, + -1.5354e-04, -8.3923e-05], + [-1.8799e-02, -1.8311e-02, -1.7822e-02, ..., -5.8889e-05, + -6.6280e-05, -1.0681e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_cross_attn.2.weight': tensor([[-0.0046, 0.0047, 0.0025, ..., 0.0187, -0.0072, -0.0151], + [-0.0025, 0.0027, 0.0030, ..., -0.0016, -0.0025, -0.0016], + [-0.0008, 0.0026, 0.0008, ..., 0.0009, -0.0027, -0.0023], + ..., + [ 0.0033, -0.0033, -0.0030, ..., 0.0014, 0.0021, -0.0142], + [ 0.0317, -0.0312, -0.0286, ..., -0.0209, 0.0284, -0.0272], + [-0.0250, 0.0240, 0.0238, ..., 0.0157, -0.0222, 0.0192]], + dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.1.weight': tensor([[ 4.0039e-02, 3.9307e-02, 3.8574e-02, ..., -2.3937e-04, + -4.1389e-04, -3.5667e-04], + [ 7.3730e-02, 7.0801e-02, 6.9824e-02, ..., -1.3447e-04, + -1.6880e-04, -2.1648e-04], + [ 3.8818e-02, 3.8086e-02, 3.7598e-02, ..., -4.3488e-04, + -4.0245e-04, -3.7003e-04], + ..., + [-4.7363e-02, -4.6143e-02, -4.5654e-02, ..., 2.0313e-04, + 9.9659e-05, 1.1873e-04], + [-2.1118e-02, -2.0752e-02, -2.0752e-02, ..., 2.8038e-04, + 2.8801e-04, 3.3951e-04], + [ 3.0151e-02, 2.9663e-02, 2.8809e-02, ..., -1.2159e-04, + -9.7752e-05, -1.9073e-04]], dtype=torch.bfloat16), 'net.blocks.4.adaln_modulation_mlp.2.weight': tensor([[ 6.3171e-03, 1.7578e-02, 5.7068e-03, ..., -8.9111e-03, + 6.8283e-04, 1.4954e-02], + [-1.4496e-03, -4.4250e-03, -4.5395e-04, ..., 1.5869e-03, + -1.2064e-04, -3.7670e-05], + [ 3.2349e-03, 4.2725e-03, 3.1433e-03, ..., -3.6774e-03, + -6.3705e-04, 1.2283e-03], + ..., + [-4.3945e-03, 4.9133e-03, -4.3030e-03, ..., 4.0894e-03, + 6.4392e-03, 1.0986e-02], + [ 1.1719e-02, 2.0142e-02, 1.1414e-02, ..., -1.3672e-02, + -7.0496e-03, -3.7384e-03], + [-1.5259e-02, -2.4048e-02, -1.4282e-02, ..., 1.6602e-02, + 1.0559e-02, 1.6327e-03]], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_proj.weight': tensor([[-0.0299, 0.0432, -0.0109, ..., -0.0097, -0.0615, -0.0374], + [ 0.0243, -0.0276, -0.0023, ..., -0.0123, 0.0186, -0.0217], + [-0.0236, -0.0035, 0.0221, ..., -0.0124, -0.0099, -0.0386], + ..., + [ 0.0047, -0.0015, -0.0108, ..., -0.0085, 0.0068, 0.0023], + [ 0.0260, 0.0103, -0.0154, ..., 0.0013, 0.0135, -0.0053], + [-0.0092, -0.0049, 0.0282, ..., -0.0035, -0.0074, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm.weight': tensor([1.6484, 1.3359, 0.8555, 1.3125, 0.6484, 0.2441, 1.1875, 0.1572, 0.1250, + 1.1250, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8789, 1.1406, + 1.2734, 1.2500, 0.2773, 0.2832, 0.3008, 0.8555, 1.4531, 0.7617, 0.5977, + 0.2119, 0.1040, 0.2285, 0.1162, 1.1094, 0.2598, 1.1172, 0.1445, 0.0869, + 1.1250, 1.1406, 0.0723, 1.0703, 1.2266, 0.3555, 0.9375, 0.2119, 1.4297, + 1.3359, 0.6055, 1.2422, 0.7578, 0.1455, 1.1797, 0.1602, 1.1953, 1.1719, + 1.1953, 0.2422, 1.2188, 1.1406, 0.1309, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1157, 0.4141, 0.4492, 0.5039, 0.7148, 0.2041, 1.1250, 0.2832, 0.8242, + 0.0393, 1.0234, 0.1934, 1.2031, 1.1641, 0.1436, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2344, 0.1357, 0.1245, 0.1875, 1.7188, 1.3516, 0.7773, 1.2969, + 0.7734, 0.2539, 0.0211, 1.1406, 0.2148, 1.2031, 1.1016, 1.0547, 1.1562, + 1.1172, 0.0142, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.2891, 0.4746, 1.4219, 0.7656, 0.1846, 0.0393, 0.2578, 1.1406, 1.0938, + 0.0654, 0.1206, 0.0287, 1.0938, 1.1719, 1.1484, 0.1777, 1.1328, 0.1582, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.k_proj.weight': tensor([[-0.0845, 0.0108, 0.0454, ..., 0.0087, -0.0420, -0.0518], + [ 0.0437, -0.0354, -0.0425, ..., -0.0066, 0.0684, -0.0474], + [-0.0276, 0.0129, 0.0167, ..., -0.0041, -0.0311, -0.0342], + ..., + [-0.0095, -0.0003, 0.0110, ..., -0.0062, 0.0020, 0.0009], + [ 0.0090, -0.0121, -0.0077, ..., 0.0059, -0.0039, 0.0009], + [ 0.0209, -0.0014, 0.0155, ..., 0.0100, 0.0125, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm.weight': tensor([1.3828, 1.3047, 0.8438, 1.2969, 0.6328, 0.2266, 1.1875, 0.1396, 0.1167, + 1.1328, 1.2266, 1.1484, 1.1172, 1.1953, 1.1797, 1.1484, 0.8750, 1.1328, + 1.2734, 1.2500, 0.2773, 0.2832, 1.1797, 1.0391, 1.3750, 0.6367, 0.5547, + 0.1914, 0.0967, 0.3359, 0.1138, 1.1094, 0.2578, 1.1172, 0.1533, 0.0791, + 1.1250, 1.1406, 0.0723, 1.0625, 1.2266, 0.3555, 0.9375, 0.4082, 1.4297, + 1.3672, 0.7266, 1.2969, 0.7734, 0.1553, 1.1875, 0.4121, 1.1875, 1.1641, + 1.1953, 0.2402, 1.2188, 1.1406, 0.1318, 0.1465, 1.0234, 1.1953, 0.9375, + 0.1147, 0.9766, 0.6055, 0.5078, 0.7383, 0.3047, 1.1172, 0.2852, 0.8242, + 0.0623, 1.0000, 0.2188, 1.1953, 1.1641, 0.1494, 1.2422, 1.1641, 1.1250, + 1.1328, 1.2266, 0.1357, 0.1250, 0.1865, 1.3047, 1.2266, 0.5352, 1.2734, + 0.8008, 0.2793, 0.0396, 1.1250, 0.2109, 1.1953, 1.1016, 1.0703, 1.1484, + 1.1172, 0.0172, 1.1641, 1.0391, 1.2422, 1.1797, 1.1797, 0.0840, 1.5938, + 0.3848, 0.2773, 1.3516, 0.7109, 0.0981, 0.0052, 0.3730, 1.1328, 1.1484, + 0.0664, 0.0806, 0.0258, 1.0859, 1.1641, 1.1484, 0.1777, 1.1328, 0.1592, + 0.0693, 1.1406], dtype=torch.bfloat16), 'net.blocks.5.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.self_attn.v_proj.weight': tensor([[-0.0004, -0.0069, -0.0187, ..., 0.0090, 0.0096, -0.0043], + [-0.0003, -0.0085, 0.0012, ..., 0.0093, 0.0048, -0.0040], + [-0.0022, -0.0310, 0.0040, ..., -0.0074, -0.0037, -0.0166], + ..., + [ 0.0065, -0.0137, -0.0095, ..., 0.0013, -0.0012, 0.0078], + [ 0.0011, -0.0033, -0.0287, ..., 0.0085, 0.0044, -0.0140], + [ 0.0071, -0.0181, 0.0354, ..., 0.0261, 0.0146, -0.0050]], + dtype=torch.bfloat16), 'net.blocks.5.self_attn.output_proj.weight': tensor([[-0.0033, 0.0089, -0.0060, ..., 0.0189, -0.0044, 0.0075], + [-0.0102, 0.0018, 0.0166, ..., -0.0084, 0.0118, 0.0181], + [-0.0115, 0.0115, 0.0110, ..., 0.0201, 0.0027, -0.0080], + ..., + [ 0.0030, 0.0005, 0.0094, ..., 0.0073, -0.0023, -0.0198], + [ 0.0203, 0.0154, -0.0234, ..., -0.0106, -0.0062, 0.0119], + [ 0.0181, 0.0082, 0.0334, ..., -0.0023, -0.0078, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_proj.weight': tensor([[-0.0111, 0.0052, 0.0046, ..., -0.0096, -0.0055, 0.0102], + [ 0.0004, 0.0047, -0.0103, ..., 0.0064, -0.0116, 0.0073], + [ 0.0297, -0.0049, -0.0032, ..., 0.0041, 0.0081, 0.0012], + ..., + [ 0.0065, -0.0112, -0.0008, ..., -0.0129, -0.0056, 0.0117], + [-0.0049, 0.0027, -0.0037, ..., -0.0008, 0.0071, 0.0001], + [-0.0031, 0.0021, 0.0045, ..., -0.0051, -0.0058, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm.weight': tensor([0.0013, 0.0154, 0.0038, 0.0413, 0.0008, 0.2383, 0.0220, 0.0022, 0.0256, + 0.0898, 0.1914, 0.0050, 0.7539, 0.3711, 0.0095, 0.0016, 0.0234, 0.0486, + 0.1504, 0.2217, 0.2217, 0.0510, 0.6836, 0.0491, 0.0530, 0.1719, 0.0039, + 0.0311, 0.6133, 0.0237, 0.0305, 0.4570, 0.0120, 0.5820, 0.0147, 0.0303, + 0.3945, 0.0503, 0.0302, 0.5156, 0.0057, 0.7383, 0.0220, 0.1104, 0.0106, + 0.2988, 0.0420, 0.2080, 0.4141, 0.0532, 0.0347, 0.0977, 0.0128, 0.1631, + 0.0203, 0.0199, 0.0469, 0.5469, 0.0018, 0.2637, 0.0021, 0.6797, 0.1104, + 0.0148, 0.0092, 0.6719, 0.0479, 0.0649, 0.4062, 0.6992, 0.0114, 0.0100, + 0.3789, 0.5586, 0.0303, 0.4121, 0.0280, 0.6328, 0.0144, 0.0093, 0.0378, + 0.0232, 0.7461, 0.7773, 0.0293, 0.0170, 0.0028, 0.1021, 0.0045, 0.0225, + 0.0122, 0.0518, 0.0164, 0.2793, 0.7773, 0.5117, 0.3105, 0.0488, 0.0312, + 0.0283, 0.0164, 0.1406, 0.0173, 0.1270, 0.0101, 0.5938, 0.6953, 0.0210, + 0.1797, 0.4902, 0.0024, 0.0028, 0.5430, 0.1118, 0.0131, 0.0211, 0.0130, + 0.0170, 0.0151, 0.0432, 0.0297, 0.1084, 0.0210, 0.6094, 0.5508, 0.0315, + 0.0114, 0.0278], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_proj.weight': tensor([[-0.0035, -0.0044, 0.0029, ..., 0.0092, -0.0028, -0.0048], + [-0.0117, -0.0005, 0.0060, ..., 0.0146, 0.0011, 0.0002], + [-0.0034, 0.0025, 0.0011, ..., 0.0078, -0.0036, 0.0121], + ..., + [ 0.0033, -0.0035, -0.0024, ..., -0.0036, 0.0049, 0.0067], + [-0.0037, -0.0024, -0.0023, ..., -0.0074, -0.0010, 0.0026], + [-0.0150, 0.0018, 0.0036, ..., -0.0116, -0.0048, -0.0066]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm.weight': tensor([1.3580e-03, 1.5320e-02, 3.6621e-03, 4.0771e-02, 7.2479e-04, 2.3828e-01, + 2.2339e-02, 2.2736e-03, 2.5757e-02, 8.9844e-02, 1.9141e-01, 4.9744e-03, + 7.5391e-01, 3.7109e-01, 9.0942e-03, 2.3346e-03, 2.3438e-02, 4.9072e-02, + 1.5039e-01, 2.2168e-01, 2.2266e-01, 5.0537e-02, 6.8359e-01, 4.8584e-02, + 5.3223e-02, 1.7188e-01, 1.6556e-03, 3.1128e-02, 6.1328e-01, 2.3682e-02, + 3.0762e-02, 4.5703e-01, 1.2207e-02, 5.8203e-01, 1.4832e-02, 3.0396e-02, + 3.9062e-01, 5.0293e-02, 3.0029e-02, 5.1562e-01, 5.7373e-03, 7.3828e-01, + 2.0142e-02, 1.1035e-01, 1.0620e-02, 3.0078e-01, 4.1992e-02, 2.0801e-01, + 4.1406e-01, 5.3223e-02, 3.4668e-02, 9.7656e-02, 1.2573e-02, 1.6309e-01, + 2.0386e-02, 1.9897e-02, 4.6875e-02, 5.4688e-01, 1.1292e-03, 2.6367e-01, + 2.2125e-03, 6.7969e-01, 1.1035e-01, 1.5625e-02, 9.2163e-03, 6.7188e-01, + 4.8096e-02, 6.4941e-02, 4.0625e-01, 6.9922e-01, 1.1353e-02, 9.9487e-03, + 3.8086e-01, 5.5859e-01, 2.8809e-02, 4.1211e-01, 2.7832e-02, 6.3281e-01, + 1.4282e-02, 1.0376e-02, 3.7842e-02, 2.2827e-02, 7.5000e-01, 7.7734e-01, + 2.9541e-02, 1.6724e-02, 2.7161e-03, 1.0205e-01, 4.4556e-03, 2.2583e-02, + 1.0742e-02, 5.1758e-02, 1.6357e-02, 2.7930e-01, 7.7734e-01, 5.0781e-01, + 3.0859e-01, 4.9316e-02, 3.1250e-02, 2.8320e-02, 1.5625e-02, 1.4062e-01, + 1.7334e-02, 1.2793e-01, 1.0132e-02, 5.9375e-01, 6.9141e-01, 2.1973e-02, + 1.7969e-01, 4.9023e-01, 2.3651e-03, 3.4027e-03, 5.4297e-01, 1.1230e-01, + 1.3123e-02, 2.0874e-02, 1.3184e-02, 1.6846e-02, 1.5259e-02, 4.2969e-02, + 2.9907e-02, 1.0840e-01, 2.0264e-02, 6.0938e-01, 5.5078e-01, 3.1494e-02, + 1.1353e-02, 2.8687e-02], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.v_proj.weight': tensor([[ 0.0009, -0.0048, -0.0214, ..., -0.0064, 0.0266, -0.0179], + [ 0.0094, -0.0109, -0.0137, ..., -0.0116, -0.0015, -0.0181], + [ 0.0255, -0.0094, 0.0036, ..., -0.0133, -0.0020, 0.0206], + ..., + [ 0.0115, -0.0004, -0.0058, ..., 0.0056, 0.0306, 0.0034], + [ 0.0149, -0.0016, 0.0123, ..., -0.0016, 0.0048, 0.0226], + [ 0.0042, -0.0050, 0.0005, ..., 0.0031, -0.0110, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.output_proj.weight': tensor([[ 0.0059, -0.0086, -0.0126, ..., 0.0058, -0.0067, 0.0132], + [-0.0022, 0.0276, -0.0303, ..., 0.0051, -0.0118, 0.0037], + [-0.0102, -0.0122, 0.0110, ..., -0.0152, 0.0162, -0.0022], + ..., + [ 0.0076, 0.0115, -0.0042, ..., -0.0029, 0.0052, 0.0284], + [-0.0156, -0.0047, -0.0166, ..., -0.0146, 0.0057, 0.0021], + [-0.0027, 0.0047, -0.0278, ..., -0.0080, 0.0010, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img.weight': tensor([[ 0.0053, 0.0131, -0.0023, ..., -0.0093, -0.0001, -0.0229], + [-0.0287, 0.0229, 0.0075, ..., -0.0137, 0.0045, 0.0222], + [ 0.0065, 0.0288, -0.0189, ..., -0.0173, 0.0311, -0.0262], + ..., + [ 0.0233, 0.0070, 0.0038, ..., -0.0023, 0.0278, 0.0064], + [-0.0109, 0.0005, 0.0151, ..., -0.0175, 0.0444, 0.0425], + [ 0.0062, -0.0068, 0.0093, ..., -0.0040, -0.0344, 0.0226]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.v_img.weight': tensor([[ 0.0059, 0.0125, 0.0190, ..., -0.0095, -0.0198, 0.0142], + [-0.0581, 0.0294, -0.0038, ..., -0.0240, 0.0066, 0.0302], + [ 0.0079, 0.0201, 0.0211, ..., 0.0215, 0.0281, -0.0259], + ..., + [-0.0017, -0.0289, 0.0537, ..., 0.0065, -0.0056, -0.0093], + [ 0.0053, -0.0080, 0.0209, ..., -0.0035, 0.0090, 0.0186], + [-0.0082, -0.0349, -0.0173, ..., -0.0400, 0.0037, -0.0457]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img.weight': tensor([[-0.0178, -0.0259, -0.0102, ..., 0.0090, -0.0137, -0.0405], + [ 0.0063, -0.0095, 0.0266, ..., -0.0400, 0.0047, 0.0391], + [ 0.0059, 0.0359, 0.0283, ..., 0.0295, -0.0233, 0.0212], + ..., + [ 0.0071, 0.0124, 0.0132, ..., -0.0139, -0.0033, 0.0095], + [ 0.0299, -0.0167, -0.0028, ..., -0.0052, -0.0032, -0.0206], + [ 0.0072, -0.0067, -0.0038, ..., -0.0032, -0.0047, 0.0527]], + dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.5.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.5.mlp.layer1.weight': tensor([[ 0.0217, -0.0303, -0.0056, ..., 0.0033, 0.0270, 0.0039], + [ 0.0339, 0.0139, -0.0090, ..., -0.0068, 0.0212, -0.0376], + [-0.0039, -0.0513, 0.0071, ..., 0.0104, -0.0204, 0.0139], + ..., + [-0.0054, -0.0160, -0.0131, ..., -0.0027, 0.0029, 0.0104], + [ 0.0023, 0.0082, 0.0008, ..., -0.0131, 0.0265, 0.0149], + [ 0.0139, -0.0126, -0.0090, ..., -0.0002, -0.0009, 0.0120]], + dtype=torch.bfloat16), 'net.blocks.5.mlp.layer2.weight': tensor([[ 0.0054, 0.0029, -0.0413, ..., 0.0046, 0.0026, 0.0177], + [-0.0212, -0.0041, -0.0210, ..., -0.0141, -0.0112, -0.0045], + [-0.0133, 0.0317, -0.0171, ..., -0.0023, -0.0041, 0.0297], + ..., + [ 0.0128, 0.0179, 0.0096, ..., -0.0132, -0.0140, -0.0057], + [ 0.0221, -0.0474, 0.0249, ..., -0.0052, 0.0312, 0.0026], + [ 0.0400, -0.0053, 0.0048, ..., -0.0050, -0.0039, 0.0293]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.1.weight': tensor([[ 3.6865e-02, 3.6133e-02, 3.4912e-02, ..., -5.0306e-05, + -3.0756e-05, -2.9683e-05], + [-4.2236e-02, -4.1504e-02, -3.9795e-02, ..., 1.5274e-06, + 1.1635e-04, 6.6757e-05], + [ 9.8145e-02, 9.6191e-02, 9.2773e-02, ..., 3.3951e-04, + 3.3188e-04, 2.9755e-04], + ..., + [ 8.7402e-02, 8.5938e-02, 8.3008e-02, ..., -3.8338e-04, + -2.9182e-04, -3.9101e-04], + [-1.0059e-01, -9.8633e-02, -9.6191e-02, ..., -3.4523e-04, + -3.8147e-04, -3.3188e-04], + [ 8.9844e-02, 8.7891e-02, 8.5449e-02, ..., -3.6240e-04, + -2.6512e-04, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_self_attn.2.weight': tensor([[-0.0010, 0.0014, -0.0231, ..., -0.0035, 0.0251, -0.0041], + [-0.0018, 0.0018, -0.0018, ..., -0.0009, 0.0045, -0.0008], + [ 0.0020, -0.0024, 0.0017, ..., 0.0085, -0.0017, 0.0091], + ..., + [ 0.0184, -0.0203, 0.0254, ..., 0.0403, -0.0131, 0.0405], + [-0.0040, 0.0023, -0.0104, ..., 0.0047, 0.0183, 0.0058], + [ 0.0133, -0.0133, 0.0171, ..., 0.0108, -0.0233, 0.0108]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.1.weight': tensor([[-3.4668e-02, -3.3691e-02, -3.2471e-02, ..., -1.2064e-04, + -1.7166e-04, -1.2493e-04], + [ 5.6641e-02, 5.4688e-02, 5.3467e-02, ..., 1.4305e-04, + 6.2943e-05, 1.0824e-04], + [ 2.5269e-02, 2.5024e-02, 2.4536e-02, ..., 1.1063e-04, + 1.0347e-04, 8.7261e-05], + ..., + [ 3.0762e-02, 2.9785e-02, 2.8809e-02, ..., 1.2684e-04, + 1.5450e-04, 1.9073e-04], + [-8.2397e-03, -7.8735e-03, -7.3853e-03, ..., -1.4019e-04, + -1.2779e-04, -4.2677e-05], + [-5.8594e-02, -5.7373e-02, -5.5664e-02, ..., 1.2302e-04, + 1.5736e-04, 3.6240e-05]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_cross_attn.2.weight': tensor([[-0.0107, 0.0187, -0.0031, ..., 0.0181, -0.0037, -0.0092], + [ 0.0030, -0.0003, 0.0012, ..., 0.0026, -0.0010, 0.0029], + [-0.0085, 0.0106, 0.0046, ..., 0.0032, -0.0006, -0.0065], + ..., + [-0.0029, 0.0164, 0.0092, ..., 0.0066, -0.0149, 0.0023], + [-0.0161, 0.0153, 0.0096, ..., -0.0018, -0.0139, -0.0228], + [-0.0120, 0.0259, -0.0039, ..., 0.0282, 0.0151, 0.0061]], + dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.1.weight': tensor([[-4.4678e-02, -4.3457e-02, -4.2725e-02, ..., 4.1008e-04, + 3.7193e-04, 3.7003e-04], + [-7.2754e-02, -7.0312e-02, -6.8359e-02, ..., 1.6117e-04, + 1.0157e-04, 1.4877e-04], + [-9.5703e-02, -9.4238e-02, -9.1309e-02, ..., 4.6158e-04, + 4.9591e-04, 5.1498e-04], + ..., + [-1.4062e-01, -1.3770e-01, -1.3477e-01, ..., -6.7353e-06, + 1.8120e-04, 5.1498e-05], + [ 8.9844e-02, 8.7402e-02, 8.6426e-02, ..., -5.3406e-04, + -5.2261e-04, -5.6076e-04], + [ 7.6172e-02, 7.5684e-02, 7.3242e-02, ..., -3.3140e-05, + -4.4346e-05, -1.0538e-04]], dtype=torch.bfloat16), 'net.blocks.5.adaln_modulation_mlp.2.weight': tensor([[ 0.0002, -0.0037, -0.0058, ..., 0.0110, 0.0066, 0.0037], + [-0.0012, 0.0004, 0.0016, ..., 0.0011, -0.0016, -0.0004], + [-0.0003, 0.0003, 0.0003, ..., 0.0019, -0.0004, -0.0003], + ..., + [ 0.0024, -0.0182, -0.0366, ..., -0.0635, 0.0376, 0.0194], + [ 0.0006, 0.0016, 0.0046, ..., 0.0168, -0.0048, -0.0017], + [-0.0051, -0.0104, -0.0159, ..., -0.0282, 0.0159, 0.0111]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_proj.weight': tensor([[ 0.0659, -0.0220, -0.0060, ..., -0.0092, -0.0206, -0.0038], + [ 0.0559, 0.0280, 0.0178, ..., 0.0050, 0.0017, 0.0161], + [-0.0684, 0.0033, -0.0040, ..., -0.0069, -0.0056, -0.0069], + ..., + [ 0.0043, -0.0046, -0.0029, ..., 0.0094, 0.0083, 0.0011], + [-0.0038, -0.0024, 0.0095, ..., 0.0001, -0.0103, -0.0061], + [ 0.0065, 0.0093, -0.0067, ..., -0.0022, 0.0015, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm.weight': tensor([ 2.1875e+00, 1.1562e+00, 1.2891e+00, 6.9922e-01, 2.8320e-01, + 8.4961e-02, 1.4297e+00, 1.3281e-01, 3.9062e-02, 3.9795e-02, + 8.3984e-01, 2.3560e-02, 4.6143e-02, 9.1016e-01, 1.0000e+00, + 4.9805e-02, 1.0620e-02, 1.5564e-02, 9.2969e-01, 2.2095e-02, + 1.2598e-01, 6.9824e-02, 8.2031e-01, 2.0938e+00, 4.9609e-01, + 3.1250e-01, 7.2266e-01, 6.5234e-01, 1.5527e-01, 7.9102e-02, + 9.1797e-01, 1.7480e-01, 2.7539e-01, 1.9141e-01, 2.5269e-02, + 1.2146e-02, 3.1738e-02, 9.1797e-01, 3.5645e-02, 8.4766e-01, + 8.4766e-01, 1.4282e-02, 3.7193e-04, 1.6562e+00, 2.0781e+00, + 1.4219e+00, 1.7031e+00, 3.8086e-01, 7.8125e-01, 6.7871e-02, + 2.2339e-02, 8.7402e-02, 8.0078e-02, 1.0742e-02, 3.4027e-03, + 1.1230e-01, 8.5156e-01, 2.7832e-02, 9.6875e-01, 9.1797e-01, + 8.9844e-01, -3.0212e-03, 3.6133e-02, 9.0234e-01, 3.5938e-01, + 1.1797e+00, 1.5547e+00, 1.6094e+00, 3.0859e-01, 1.2695e-01, + 4.1016e-01, 5.8594e-01, 1.8311e-02, 6.9531e-01, 8.7891e-01, + 8.4766e-01, 9.1797e-01, 9.7656e-03, 3.7695e-01, 1.1328e+00, + 2.1851e-02, 6.3477e-02, 3.9795e-02, 9.7266e-01, 2.6855e-02, + 9.5312e-01, 1.8594e+00, 6.9141e-01, 8.5547e-01, 4.5703e-01, + 1.4141e+00, 2.3340e-01, 1.5723e-01, 7.1289e-02, 2.5195e-01, + 8.5938e-01, 6.1768e-02, 8.9844e-01, 6.6797e-01, 9.2969e-01, + 8.7891e-01, 1.0693e-01, 8.6719e-01, 8.3203e-01, 2.3071e-02, + 1.9043e-02, 6.5308e-03, 9.9609e-01, 3.0273e-01, 1.1406e+00, + 9.0625e-01, 3.1055e-01, 3.0664e-01, 5.8105e-02, 2.8687e-02, + 5.3223e-02, 2.0020e-01, -1.5831e-04, 1.5430e-01, 7.8125e-01, + 8.4961e-02, 3.8281e-01, 5.6885e-02, 1.3428e-02, 8.5156e-01, + 2.3828e-01, 4.4434e-02, 9.3750e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.k_proj.weight': tensor([[-0.0376, -0.0035, 0.0124, ..., 0.0101, 0.0066, 0.0036], + [-0.0349, 0.0007, 0.0023, ..., 0.0050, 0.0129, 0.0063], + [-0.0786, -0.0091, -0.0063, ..., -0.0203, 0.0212, -0.0199], + ..., + [ 0.0101, 0.0068, 0.0168, ..., -0.0042, -0.0096, -0.0035], + [ 0.0078, -0.0034, -0.0049, ..., 0.0044, -0.0021, 0.0058], + [-0.0108, -0.0118, 0.0140, ..., -0.0210, -0.0234, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm.weight': tensor([1.3750e+00, 1.8281e+00, 1.3125e+00, 7.4609e-01, 2.6758e-01, 7.9590e-02, + 1.4297e+00, 1.3770e-01, 3.7354e-02, 2.0264e-02, 8.4375e-01, 1.4343e-02, + 5.4688e-02, 9.1016e-01, 1.0000e+00, 2.6611e-02, 1.0132e-02, 1.5991e-02, + 9.2969e-01, 2.0264e-02, 1.2402e-01, 6.8359e-02, 1.2891e+00, 1.5859e+00, + 7.3047e-01, 5.1953e-01, 6.5625e-01, 5.5469e-01, 1.5625e-01, 3.4180e-02, + 9.1406e-01, 1.0205e-01, 2.7539e-01, 1.9336e-01, 3.0518e-02, 2.6978e-02, + 1.4404e-02, 9.1797e-01, 3.1128e-02, 8.4766e-01, 8.4766e-01, 1.4221e-02, + 4.0245e-04, 1.6562e+00, 1.2422e+00, 1.4141e+00, 1.9688e+00, 3.6719e-01, + 7.8516e-01, 7.0312e-02, 8.0566e-03, 4.6631e-02, 3.5156e-02, 1.1169e-02, + 1.0071e-02, 1.2012e-01, 8.5156e-01, 2.0874e-02, 9.6875e-01, 9.2188e-01, + 8.9844e-01, 4.3640e-03, 3.7598e-02, 9.0234e-01, 1.4297e+00, 9.3359e-01, + 1.4531e+00, 1.5469e+00, 3.1641e-01, 1.2500e-01, 4.8828e-01, 5.8594e-01, + 2.0264e-02, 6.9531e-01, 8.8281e-01, 8.4766e-01, 9.2188e-01, 2.0874e-02, + 3.7891e-01, 1.1328e+00, 2.1484e-02, 6.2988e-02, 4.2236e-02, 9.7656e-01, + 2.6489e-02, 9.5312e-01, 1.7344e+00, 1.2969e+00, 6.7969e-01, 3.3594e-01, + 1.5156e+00, 4.5312e-01, 1.7676e-01, 9.6191e-02, 3.3398e-01, 8.5938e-01, + 3.6865e-02, 9.0234e-01, 6.6797e-01, 9.2578e-01, 8.7500e-01, 1.0986e-01, + 8.6719e-01, 8.3203e-01, 2.6855e-02, 1.9653e-02, 6.4697e-03, 1.1016e+00, + 1.2344e+00, 1.0781e+00, 8.2812e-01, 4.2578e-01, 5.2344e-01, 4.5898e-02, + 3.1250e-02, 8.7402e-02, 2.0801e-01, 7.1335e-04, 1.5430e-01, 7.8125e-01, + 7.4219e-02, 3.8281e-01, 6.0791e-02, 1.5503e-02, 8.4766e-01, 2.3730e-01, + 4.7119e-02, 9.4141e-01], dtype=torch.bfloat16), 'net.blocks.6.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.self_attn.v_proj.weight': tensor([[ 0.0004, 0.0033, 0.0073, ..., -0.0079, -0.0045, 0.0093], + [-0.0022, 0.0111, -0.0142, ..., -0.0039, 0.0125, 0.0026], + [-0.0030, 0.0008, -0.0096, ..., -0.0088, -0.0033, 0.0156], + ..., + [-0.0175, -0.0117, -0.0009, ..., -0.0012, -0.0234, 0.0123], + [ 0.0007, -0.0116, -0.0007, ..., 0.0047, -0.0170, -0.0063], + [-0.0023, -0.0164, 0.0135, ..., 0.0128, -0.0056, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.6.self_attn.output_proj.weight': tensor([[-0.0117, -0.0113, -0.0135, ..., -0.0170, 0.0049, -0.0253], + [ 0.0055, 0.0042, 0.0047, ..., 0.0070, 0.0096, 0.0057], + [-0.0082, 0.0052, -0.0024, ..., 0.0015, 0.0162, 0.0085], + ..., + [-0.0034, -0.0087, 0.0071, ..., 0.0069, 0.0129, 0.0179], + [ 0.0066, -0.0017, -0.0089, ..., -0.0303, -0.0106, 0.0009], + [-0.0178, -0.0197, 0.0145, ..., 0.0193, -0.0135, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_proj.weight': tensor([[-0.0006, -0.0089, 0.0128, ..., -0.0027, 0.0019, 0.0018], + [ 0.0023, 0.0122, 0.0057, ..., 0.0036, 0.0045, 0.0035], + [-0.0109, 0.0405, -0.0018, ..., -0.0164, 0.0011, -0.0050], + ..., + [ 0.0044, 0.0036, -0.0078, ..., -0.0113, 0.0018, -0.0025], + [-0.0014, 0.0093, -0.0042, ..., -0.0023, -0.0034, 0.0027], + [ 0.0028, 0.0005, 0.0056, ..., -0.0072, -0.0135, 0.0089]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm.weight': tensor([0.0330, 0.1934, 0.5977, 0.1230, 0.7383, 0.0752, 0.0488, 0.1787, 0.0247, + 0.4961, 0.5625, 0.0295, 0.0339, 0.0688, 0.1738, 0.1167, 0.2969, 0.0549, + 0.3750, 0.0281, 0.2197, 0.1318, 0.0334, 0.0060, 0.0427, 0.5586, 0.0292, + 0.0894, 0.0835, 0.1816, 0.0269, 0.0198, 0.0371, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0298, 0.3594, 0.0398, 0.4883, 0.5977, 0.0371, 0.0535, 0.0483, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0293, 0.6094, 0.1670, 0.1128, 0.1001, + 0.0122, 0.7344, 0.0371, 0.7812, 0.2031, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0552, 0.1270, 0.6992, 0.5859, 0.5352, 0.0325, 0.0195, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0525, 0.0708, 0.0240, 0.0483, 0.0898, 0.0327, + 0.0170, 0.0415, 0.5977, 0.0454, 0.7539, 0.0586, 0.0248, 0.0253, 0.1299, + 0.0312, 0.5898, 0.0096, 0.0474, 0.6758, 0.0464, 0.6719, 0.0222, 0.2441, + 0.0280, 0.0576, 0.0986, 0.2158, 0.0215, 0.0591, 0.0359, 0.5625, 0.0518, + 0.0178, 0.0205, 0.5078, 0.7109, 0.1230, 0.6016, 0.0217, 0.3164, 0.7305, + 0.5703, 0.0184, 0.6445, 0.6562, 0.7148, 0.0288, 0.4219, 0.0400, 0.0276, + 0.0361, 0.6250], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_proj.weight': tensor([[-0.0015, -0.0011, 0.0040, ..., 0.0028, -0.0073, -0.0085], + [ 0.0079, -0.0006, -0.0033, ..., -0.0042, -0.0093, 0.0104], + [ 0.0221, -0.0044, 0.0052, ..., 0.0095, 0.0042, 0.0018], + ..., + [-0.0063, -0.0023, -0.0080, ..., 0.0086, 0.0017, 0.0015], + [-0.0015, 0.0035, -0.0049, ..., 0.0055, -0.0017, -0.0038], + [ 0.0145, -0.0102, 0.0054, ..., -0.0195, 0.0081, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm.weight': tensor([0.0330, 0.1924, 0.5977, 0.1216, 0.7383, 0.0752, 0.0483, 0.1797, 0.0248, + 0.4961, 0.5625, 0.0298, 0.0339, 0.0688, 0.1738, 0.1167, 0.2988, 0.0557, + 0.3750, 0.0282, 0.2197, 0.1318, 0.0334, 0.0059, 0.0427, 0.5586, 0.0288, + 0.0889, 0.0835, 0.1816, 0.0266, 0.0199, 0.0374, 0.0332, 0.2158, 0.7188, + 0.2256, 0.0299, 0.3594, 0.0388, 0.4883, 0.5977, 0.0369, 0.0537, 0.0488, + 0.0796, 0.2754, 0.1348, 0.3730, 0.0292, 0.6094, 0.1660, 0.1128, 0.1006, + 0.0122, 0.7344, 0.0371, 0.7852, 0.2051, 0.7461, 0.1187, 0.1709, 0.0186, + 0.0549, 0.1279, 0.6992, 0.5859, 0.5352, 0.0322, 0.0194, 0.0229, 0.0583, + 0.2539, 0.1406, 0.0303, 0.0520, 0.0713, 0.0242, 0.0483, 0.0898, 0.0325, + 0.0172, 0.0417, 0.5977, 0.0444, 0.7539, 0.0586, 0.0249, 0.0254, 0.1299, + 0.0312, 0.5898, 0.0097, 0.0479, 0.6758, 0.0464, 0.6719, 0.0217, 0.2441, + 0.0277, 0.0576, 0.1001, 0.2158, 0.0209, 0.0591, 0.0359, 0.5625, 0.0522, + 0.0179, 0.0203, 0.5078, 0.7070, 0.1230, 0.6016, 0.0227, 0.3164, 0.7266, + 0.5703, 0.0184, 0.6445, 0.6523, 0.7148, 0.0288, 0.4219, 0.0400, 0.0278, + 0.0361, 0.6211], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.v_proj.weight': tensor([[ 0.0002, -0.0077, 0.0036, ..., -0.0005, -0.0069, -0.0099], + [-0.0017, -0.0019, -0.0089, ..., -0.0102, -0.0156, 0.0090], + [-0.0034, 0.0004, -0.0025, ..., -0.0093, 0.0025, -0.0040], + ..., + [-0.0018, -0.0072, -0.0098, ..., -0.0048, -0.0024, 0.0212], + [ 0.0150, 0.0006, -0.0117, ..., -0.0014, 0.0171, -0.0054], + [ 0.0037, -0.0001, 0.0003, ..., 0.0038, 0.0275, 0.0075]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.output_proj.weight': tensor([[-0.0059, -0.0087, 0.0165, ..., -0.0073, 0.0012, 0.0029], + [ 0.0179, 0.0107, -0.0092, ..., 0.0098, -0.0098, 0.0009], + [ 0.0077, 0.0068, -0.0044, ..., 0.0099, -0.0092, 0.0154], + ..., + [ 0.0094, -0.0110, -0.0035, ..., 0.0034, 0.0066, 0.0237], + [-0.0151, -0.0104, 0.0061, ..., -0.0020, -0.0275, -0.0025], + [ 0.0053, 0.0226, 0.0157, ..., -0.0018, -0.0043, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img.weight': tensor([[ 0.0356, 0.0056, 0.0116, ..., -0.0032, -0.0211, -0.0064], + [-0.0101, -0.0284, -0.0192, ..., 0.0376, 0.0210, 0.0229], + [ 0.0139, -0.0278, 0.0116, ..., 0.0396, -0.0061, -0.0069], + ..., + [ 0.0029, 0.0034, -0.0311, ..., 0.0344, 0.0062, -0.0125], + [-0.0236, -0.0104, -0.0061, ..., 0.0332, -0.0117, -0.0173], + [ 0.0231, 0.0405, 0.0107, ..., -0.0281, -0.0138, 0.0366]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.v_img.weight': tensor([[ 0.0014, 0.0166, 0.0006, ..., 0.0273, -0.0352, 0.0152], + [-0.0510, -0.0082, 0.0398, ..., 0.0157, 0.0260, -0.0286], + [ 0.0272, -0.0182, -0.0074, ..., -0.0388, 0.0004, 0.0398], + ..., + [-0.0160, -0.0195, -0.0021, ..., -0.0015, -0.0051, 0.0033], + [-0.0187, 0.0042, 0.0098, ..., 0.0287, -0.0120, -0.0101], + [-0.0181, 0.0039, 0.0247, ..., -0.0149, 0.0066, -0.0244]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img.weight': tensor([[ 0.0085, 0.0096, 0.0140, ..., -0.0112, -0.0023, -0.0253], + [-0.0166, 0.0330, -0.0171, ..., 0.0024, -0.0403, -0.0110], + [ 0.0086, 0.0347, 0.0275, ..., 0.0153, 0.0068, 0.0088], + ..., + [ 0.0271, -0.0168, -0.0105, ..., -0.0101, 0.0322, 0.0381], + [-0.0281, 0.0454, -0.0058, ..., -0.0040, 0.0006, -0.0243], + [-0.0083, 0.0581, 0.0061, ..., 0.0147, 0.0239, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.6.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.6.mlp.layer1.weight': tensor([[ 2.7008e-03, -6.2256e-03, 3.0029e-02, ..., 1.7929e-03, + -3.0640e-02, -5.0659e-03], + [ 1.2512e-02, 3.8300e-03, -4.1199e-03, ..., -3.8338e-04, + 7.4463e-03, -7.4768e-03], + [ 6.5613e-03, 6.9618e-05, 1.3611e-02, ..., 1.3000e-02, + -1.9775e-02, 1.7853e-03], + ..., + [-1.9897e-02, 5.9509e-03, 2.3956e-03, ..., -2.7344e-02, + 8.9111e-03, -2.0508e-02], + [ 5.4321e-03, 4.5471e-03, 1.7456e-02, ..., -9.0332e-03, + -1.1963e-02, -1.3733e-03], + [ 2.1851e-02, -1.4954e-02, 1.4343e-02, ..., 7.0496e-03, + -2.6398e-03, 1.4648e-02]], dtype=torch.bfloat16), 'net.blocks.6.mlp.layer2.weight': tensor([[-0.0066, -0.0115, 0.0024, ..., 0.0168, -0.0134, 0.0104], + [ 0.0156, -0.0120, 0.0004, ..., -0.0113, -0.0278, -0.0009], + [-0.0280, 0.0042, -0.0062, ..., -0.0069, 0.0018, -0.0059], + ..., + [ 0.0101, -0.0146, 0.0072, ..., 0.0017, -0.0027, -0.0041], + [-0.0148, 0.0258, -0.0017, ..., -0.0062, -0.0049, -0.0031], + [-0.0022, 0.0035, -0.0048, ..., -0.0097, 0.0222, 0.0040]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.1.weight': tensor([[-6.1768e-02, -6.0547e-02, -5.9326e-02, ..., -1.5259e-04, + -1.2016e-04, -9.4414e-05], + [-5.7373e-02, -5.5908e-02, -5.4932e-02, ..., 5.9891e-04, + 6.7520e-04, 6.5613e-04], + [ 3.3447e-02, 3.2227e-02, 3.1128e-02, ..., -1.1749e-03, + -1.1826e-03, -1.1673e-03], + ..., + [ 4.8340e-02, 4.7607e-02, 4.6143e-02, ..., -1.7090e-03, + -1.5488e-03, -1.6327e-03], + [ 4.0039e-02, 3.9062e-02, 3.8086e-02, ..., 2.4033e-04, + 2.5558e-04, 2.0599e-04], + [ 1.7700e-02, 1.6968e-02, 1.6724e-02, ..., 4.0436e-04, + 3.6621e-04, 4.1389e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_self_attn.2.weight': tensor([[ 1.1658e-02, 3.5889e-02, -1.6113e-02, ..., -2.5757e-02, + -2.6703e-03, 4.9744e-03], + [-4.4632e-04, -1.4801e-03, -1.8024e-04, ..., -1.2131e-03, + 3.1662e-04, -4.1485e-05], + [ 1.1215e-03, 4.2725e-03, 3.1662e-04, ..., -2.0695e-04, + -9.4604e-04, -2.9373e-04], + ..., + [ 1.2756e-02, -1.3733e-02, 2.6611e-02, ..., 3.0273e-02, + -1.9409e-02, -2.2583e-02], + [ 4.5967e-04, -5.7373e-03, 9.0332e-03, ..., 1.1169e-02, + -2.9297e-03, -5.6458e-03], + [-1.8311e-02, -8.0566e-03, -1.9287e-02, ..., -1.8311e-02, + 1.8921e-02, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.1.weight': tensor([[-8.0566e-03, -7.9956e-03, -8.1177e-03, ..., 1.6499e-04, + 2.6131e-04, 8.6308e-05], + [-2.1667e-03, 1.5869e-03, -8.3160e-04, ..., 1.7834e-04, + 8.4400e-05, 2.0504e-04], + [ 2.9663e-02, 2.8931e-02, 2.8320e-02, ..., 2.5177e-04, + 1.5545e-04, 1.9264e-04], + ..., + [ 3.2471e-02, 3.1738e-02, 3.1128e-02, ..., -1.1969e-04, + -2.6703e-04, -2.0313e-04], + [ 5.9814e-02, 5.8350e-02, 5.7373e-02, ..., -1.4877e-04, + -1.1778e-04, -1.6880e-04], + [ 5.7617e-02, 5.5908e-02, 5.4443e-02, ..., -1.7548e-04, + -1.9264e-04, -1.3065e-04]], dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0042, 0.0039, -0.0156, ..., -0.0045, 0.0001, 0.0150], + [-0.0015, -0.0015, -0.0032, ..., 0.0016, -0.0045, 0.0015], + [-0.0013, -0.0014, 0.0021, ..., 0.0018, -0.0037, -0.0020], + ..., + [ 0.0042, 0.0042, 0.0020, ..., 0.0075, -0.0003, 0.0137], + [-0.0159, -0.0160, 0.0164, ..., 0.0077, 0.0137, 0.0011], + [-0.0029, -0.0029, 0.0076, ..., -0.0104, -0.0003, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.1.weight': tensor([[-0.0737, -0.0713, -0.0698, ..., 0.0007, 0.0006, 0.0005], + [-0.0097, -0.0098, -0.0096, ..., 0.0004, 0.0003, 0.0003], + [-0.1504, -0.1465, -0.1436, ..., 0.0004, 0.0003, 0.0003], + ..., + [ 0.0566, 0.0554, 0.0542, ..., -0.0004, -0.0004, -0.0004], + [ 0.0601, 0.0588, 0.0576, ..., -0.0004, -0.0005, -0.0005], + [ 0.0864, 0.0850, 0.0830, ..., -0.0006, -0.0006, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.6.adaln_modulation_mlp.2.weight': tensor([[-0.0135, 0.0036, 0.0059, ..., 0.0013, 0.0033, 0.0110], + [-0.0002, -0.0029, -0.0007, ..., 0.0007, 0.0004, 0.0003], + [ 0.0031, 0.0050, -0.0003, ..., -0.0036, -0.0035, -0.0030], + ..., + [-0.0312, 0.0167, 0.0121, ..., -0.0023, -0.0010, 0.0264], + [ 0.0027, 0.0007, 0.0029, ..., -0.0024, -0.0025, -0.0032], + [-0.0201, -0.0044, -0.0359, ..., 0.0099, 0.0111, 0.0208]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_proj.weight': tensor([[ 0.0103, -0.0088, -0.0104, ..., 0.0027, -0.0140, -0.0064], + [ 0.0171, 0.0273, -0.0199, ..., 0.0115, -0.0125, -0.0002], + [-0.0087, -0.0056, -0.0087, ..., 0.0046, -0.0233, 0.0099], + ..., + [ 0.0264, 0.0080, -0.0197, ..., -0.0129, 0.0018, 0.0123], + [ 0.0134, 0.0053, -0.0016, ..., -0.0085, 0.0160, 0.0062], + [ 0.0039, -0.0043, -0.0128, ..., 0.0076, -0.0104, 0.0211]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm.weight': tensor([1.7266, 1.7266, 0.5938, 0.9297, 0.4121, 0.5977, 1.2266, 0.4336, 0.2500, + 0.0532, 0.1670, 0.0281, 0.2080, 0.1030, 0.5977, 0.0913, 1.0391, 0.1543, + 0.8086, 0.3301, 0.1021, 0.4219, 1.7344, 1.6094, 1.4609, 0.4570, 1.3984, + 0.7070, 1.2969, 1.2344, 0.2246, 0.5273, 0.1484, 0.3379, 0.0461, 0.0825, + 0.0796, 0.2021, 0.0229, 0.4551, 0.1426, 0.1318, 0.0038, 0.8242, 1.1484, + 0.6953, 1.1953, 0.8984, 0.7930, 0.8242, 0.3750, 0.0869, 0.0569, 0.1631, + 0.0352, 0.3203, 0.6875, 0.4805, 0.1377, 1.0391, 0.1836, 0.7070, 0.6641, + 0.7305, 0.7461, 0.4746, 1.4609, 1.0938, 0.4082, 0.2158, 0.3047, 0.0996, + 1.2109, 0.0796, 1.0156, 0.0654, 1.0391, 0.0040, 0.0840, 0.0801, 0.0062, + 0.2324, 1.0000, 0.1040, 0.0596, 0.1709, 0.8359, 1.0781, 0.8906, 1.4141, + 1.2812, 0.3184, 0.4941, 0.5742, 0.1050, 0.0549, 0.0349, 0.4199, 0.0713, + 0.9922, 0.0349, 0.0466, 0.1699, 0.0654, 0.0413, 0.7734, 0.0294, 1.7266, + 1.6484, 1.4688, 1.5312, 1.5625, 0.5078, 0.5000, 0.4668, 0.2480, 0.0525, + 1.0938, 0.4082, 0.6406, 0.9805, 0.1338, 0.1187, 0.2158, 0.1348, 1.0625, + 0.0835, 0.5117], dtype=torch.bfloat16), 'net.blocks.7.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.k_proj.weight': tensor([[ 0.0413, 0.0113, 0.0021, ..., 0.0217, -0.0327, -0.0085], + [ 0.0369, 0.0131, -0.0050, ..., 0.0056, -0.0139, -0.0058], + [-0.0145, -0.0068, -0.0150, ..., -0.0214, -0.0054, -0.0042], + ..., + [-0.0032, 0.0049, -0.0269, ..., -0.0028, -0.0245, -0.0233], + [ 0.0144, -0.0190, -0.0110, ..., 0.0131, 0.0110, 0.0023], + [ 0.0012, 0.0123, 0.0083, ..., 0.0034, -0.0107, 0.0026]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm.weight': tensor([ 1.7969, 1.5469, 0.8477, 1.1016, 0.4844, 0.6250, 1.2266, 0.4336, + 0.4062, 0.0522, 0.2256, 0.0267, 0.2227, 0.1030, 0.6016, 0.0913, + 1.0391, 0.1543, 0.8086, 0.3301, 0.1021, 0.4219, 1.5938, 1.8047, + 1.4219, 0.6445, 1.3828, 0.7617, 1.3047, 1.2422, 0.2305, 0.5273, + 0.1494, 0.3398, 0.0469, 0.0688, 0.0786, 0.2021, 0.0233, 0.4551, + 0.1436, 0.1309, 0.0030, 1.4062, 0.9180, 0.7969, 1.3281, 0.9062, + 0.7734, 0.8594, 0.3086, 0.0635, 0.0630, 0.2432, 0.0347, 0.3164, + 0.6875, 0.4766, 0.1377, 1.0391, 0.1826, 0.7070, 0.6641, 0.7305, + 0.7070, 0.9609, 1.4062, 0.8594, 0.3535, 0.1250, 0.4434, 0.1230, + 1.2031, 0.0801, 1.0156, 0.0654, 1.0391, 0.0022, 0.0859, 0.0806, + -0.0090, 0.2314, 1.0000, 0.1040, 0.0593, 0.1709, 1.3672, 0.8125, + 0.9805, 1.3594, 1.2891, 0.2852, 0.5039, 0.4922, 0.0928, 0.0713, + 0.0432, 0.4219, 0.0718, 0.9883, 0.0369, 0.0476, 0.1709, 0.0645, + 0.0417, 0.7695, 0.0300, 1.4844, 1.7188, 1.4844, 1.4141, 1.3984, + 0.5625, 0.4980, 0.5117, 0.2520, 0.0486, 1.0938, 0.4082, 0.6406, + 0.9844, 0.1328, 0.1187, 0.2129, 0.1348, 1.0625, 0.0830, 0.5117], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.self_attn.v_proj.weight': tensor([[ 0.0071, -0.0065, 0.0179, ..., -0.0015, 0.0025, 0.0155], + [ 0.0017, -0.0082, -0.0018, ..., -0.0069, -0.0184, 0.0104], + [-0.0037, 0.0071, -0.0049, ..., 0.0057, 0.0026, -0.0194], + ..., + [-0.0109, -0.0214, -0.0155, ..., -0.0052, -0.0038, -0.0205], + [-0.0037, 0.0204, 0.0243, ..., -0.0425, -0.0110, -0.0025], + [-0.0031, 0.0259, 0.0123, ..., -0.0145, 0.0147, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.7.self_attn.output_proj.weight': tensor([[-0.0038, 0.0062, -0.0103, ..., -0.0129, 0.0425, -0.0028], + [-0.0130, 0.0041, -0.0079, ..., 0.0036, -0.0062, -0.0208], + [ 0.0094, -0.0208, 0.0164, ..., -0.0148, 0.0014, -0.0118], + ..., + [ 0.0325, -0.0049, 0.0065, ..., 0.0160, -0.0126, -0.0195], + [ 0.0036, 0.0210, 0.0272, ..., 0.0085, -0.0278, 0.0050], + [ 0.0041, -0.0320, 0.0154, ..., 0.0216, 0.0095, -0.0099]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_proj.weight': tensor([[-0.0033, 0.0052, 0.0019, ..., -0.0081, 0.0008, 0.0099], + [-0.0089, -0.0032, -0.0004, ..., 0.0142, 0.0148, -0.0101], + [-0.0075, -0.0016, -0.0023, ..., -0.0085, -0.0015, -0.0070], + ..., + [-0.0029, 0.0083, -0.0083, ..., 0.0038, 0.0022, 0.0029], + [ 0.0056, 0.0073, -0.0015, ..., 0.0047, 0.0054, -0.0034], + [ 0.0005, 0.0008, -0.0006, ..., 0.0084, 0.0103, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm.weight': tensor([0.3398, 0.1182, 0.0311, 0.0359, 0.0178, 0.0192, 0.7578, 0.0786, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0635, 0.7031, 0.2949, 0.0422, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0532, 0.1143, 0.0435, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0811, 0.5820, 0.0762, + 0.4707, 0.0530, 0.6250, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1680, 0.0679, 0.1689, 0.1143, 0.7266, + 0.0327, 0.0303, 0.7734, 0.2158, 0.0864, 0.0251, 0.0145, 0.0630, 0.0820, + 0.1196, 0.0884, 0.0442, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0825, + 0.7109, 0.1104, 0.6953, 0.7461, 0.0557, 0.3594, 0.0464, 0.7852, 0.1426, + 0.7188, 0.1289, 0.0277, 0.1138, 0.0248, 0.0962, 0.0309, 0.5352, 0.1230, + 0.0240, 0.6445, 0.0942, 0.2793, 0.1602, 0.0153, 0.4707, 0.4902, 0.6992, + 0.0179, 0.2578, 0.0265, 0.7031, 0.0122, 0.0947, 0.2598, 0.2109, 0.0339, + 0.0398, 0.6992, 0.7188, 0.6211, 0.0923, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1484], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_proj.weight': tensor([[ 1.4648e-02, 3.7384e-03, 1.5198e-02, ..., 2.7084e-04, + -1.2024e-02, 1.4801e-03], + [-2.9602e-03, -1.5259e-03, -7.2632e-03, ..., -4.5776e-03, + 6.1951e-03, -1.5991e-02], + [-3.0060e-03, -6.0081e-05, -1.2146e-02, ..., -3.5400e-03, + -3.3875e-03, -1.1841e-02], + ..., + [-8.4839e-03, -3.3722e-03, 2.1667e-03, ..., 4.1809e-03, + 1.4709e-02, -5.1880e-03], + [-2.0752e-03, 2.9755e-03, -6.1646e-03, ..., -1.9989e-03, + 1.3611e-02, -1.2329e-02], + [ 2.2984e-04, 5.9814e-03, -1.4221e-02, ..., 4.9133e-03, + 2.3651e-03, -7.5989e-03]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm.weight': tensor([0.3398, 0.1172, 0.0306, 0.0352, 0.0179, 0.0192, 0.7539, 0.0796, 0.0542, + 0.0752, 0.0869, 0.5938, 0.1494, 0.3203, 0.1108, 0.6562, 0.6836, 0.6992, + 0.0522, 0.0659, 0.5977, 0.3711, 0.0640, 0.7031, 0.2969, 0.0427, 0.7148, + 0.7148, 0.6484, 0.6250, 0.0386, 0.8633, 0.2637, 0.0530, 0.1143, 0.0432, + 0.0566, 0.1416, 0.5586, 0.7656, 0.6289, 0.3008, 0.0815, 0.5820, 0.0757, + 0.4727, 0.0520, 0.6289, 0.7422, 0.6484, 0.2412, 0.3359, 0.2637, 0.2520, + 0.5977, 0.2051, 0.7148, 0.6758, 0.1670, 0.0684, 0.1689, 0.1138, 0.7266, + 0.0330, 0.0298, 0.7734, 0.2158, 0.0869, 0.0248, 0.0143, 0.0630, 0.0820, + 0.1206, 0.0879, 0.0444, 0.6250, 0.0791, 0.3789, 0.3633, 0.6094, 0.0815, + 0.7109, 0.1108, 0.6953, 0.7461, 0.0554, 0.3594, 0.0466, 0.7852, 0.1426, + 0.7227, 0.1289, 0.0273, 0.1128, 0.0245, 0.0962, 0.0304, 0.5352, 0.1216, + 0.0242, 0.6445, 0.0938, 0.2793, 0.1592, 0.0153, 0.4688, 0.4902, 0.6992, + 0.0182, 0.2578, 0.0266, 0.7031, 0.0126, 0.0952, 0.2598, 0.2109, 0.0332, + 0.0396, 0.6992, 0.7188, 0.6211, 0.0918, 0.7344, 0.3984, 0.8125, 0.2021, + 0.0265, 0.1475], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.v_proj.weight': tensor([[ 1.4465e-02, 1.1520e-03, -1.2268e-02, ..., 1.0610e-05, + 7.3853e-03, 1.0315e-02], + [ 6.2256e-03, -5.4932e-04, -5.8594e-03, ..., 1.0223e-03, + -1.5259e-02, 4.9133e-03], + [ 1.5747e-02, -5.1880e-03, -8.7280e-03, ..., 2.3651e-03, + 4.4250e-03, 5.5237e-03], + ..., + [-1.9073e-03, -7.8735e-03, -8.5449e-03, ..., 1.1215e-03, + -2.0599e-03, 1.1414e-02], + [ 1.6357e-02, 5.7983e-03, -2.2430e-03, ..., -3.9062e-03, + 5.0964e-03, 2.3438e-02], + [ 7.5073e-03, -9.9945e-04, 1.0681e-02, ..., 3.2349e-03, + -1.7822e-02, 2.6611e-02]], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.output_proj.weight': tensor([[-0.0022, -0.0041, -0.0034, ..., -0.0231, 0.0239, 0.0172], + [-0.0079, -0.0004, -0.0086, ..., 0.0106, -0.0168, 0.0091], + [ 0.0121, 0.0276, 0.0031, ..., 0.0260, -0.0148, -0.0025], + ..., + [ 0.0058, 0.0018, -0.0238, ..., 0.0042, 0.0111, 0.0072], + [ 0.0083, 0.0104, -0.0036, ..., 0.0214, -0.0053, -0.0045], + [ 0.0053, 0.0074, -0.0018, ..., -0.0066, -0.0046, 0.0010]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img.weight': tensor([[-0.0124, -0.0410, 0.0128, ..., -0.0254, -0.0033, -0.0173], + [ 0.0476, 0.0315, 0.0141, ..., -0.0148, -0.0069, 0.0049], + [-0.0096, -0.0270, 0.0089, ..., 0.0332, 0.0148, -0.0081], + ..., + [-0.0051, -0.0019, 0.0162, ..., -0.0177, 0.0171, -0.0132], + [ 0.0052, -0.0187, -0.0068, ..., 0.0248, -0.0142, -0.0076], + [-0.0143, -0.0047, 0.0150, ..., -0.0028, -0.0015, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.v_img.weight': tensor([[-0.0317, -0.0041, 0.0099, ..., 0.0255, -0.0253, -0.0129], + [-0.0042, -0.0139, -0.0160, ..., -0.0108, -0.0403, -0.0408], + [ 0.0175, -0.0229, -0.0173, ..., 0.0525, 0.0197, 0.0276], + ..., + [ 0.0113, 0.0052, 0.0322, ..., -0.0183, -0.0053, 0.0024], + [ 0.0053, -0.0057, 0.0162, ..., 0.0043, 0.0255, 0.0043], + [ 0.0270, 0.0085, 0.0159, ..., 0.0189, 0.0145, -0.0327]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img.weight': tensor([[-0.0259, -0.0505, -0.0075, ..., -0.0532, 0.0104, 0.0029], + [-0.0240, -0.0215, -0.0106, ..., -0.0050, 0.0097, -0.0168], + [ 0.0077, -0.0034, -0.0061, ..., -0.0488, 0.0165, 0.0308], + ..., + [ 0.0071, 0.0172, -0.0106, ..., -0.0300, 0.0005, 0.0435], + [ 0.0052, -0.0261, 0.0245, ..., 0.0190, -0.0308, 0.0066], + [-0.0167, 0.0125, 0.0102, ..., -0.0247, 0.0120, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.7.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.7.mlp.layer1.weight': tensor([[-3.9673e-03, -9.2773e-03, 2.7222e-02, ..., 7.7209e-03, + 1.7471e-03, -5.0545e-05], + [ 1.1047e-02, -1.0986e-02, 8.6060e-03, ..., 5.2185e-03, + 9.4604e-03, 2.2705e-02], + [ 1.6357e-02, -2.1057e-03, -3.9368e-03, ..., 3.7994e-03, + -2.0020e-02, 1.7456e-02], + ..., + [-1.3733e-02, 4.3945e-03, 2.5024e-03, ..., 3.6011e-03, + 1.9455e-03, 1.7212e-02], + [ 6.9885e-03, -2.7344e-02, -1.1780e-02, ..., -3.1433e-03, + 1.6724e-02, -1.1597e-02], + [ 9.0942e-03, -1.7578e-02, 1.7090e-02, ..., 2.3560e-02, + 1.5320e-02, 8.5449e-03]], dtype=torch.bfloat16), 'net.blocks.7.mlp.layer2.weight': tensor([[-0.0034, 0.0172, 0.0021, ..., 0.0089, -0.0062, 0.0170], + [-0.0155, 0.0100, -0.0066, ..., -0.0019, 0.0017, -0.0061], + [-0.0041, -0.0052, -0.0219, ..., 0.0167, 0.0101, 0.0129], + ..., + [ 0.0167, -0.0270, -0.0108, ..., 0.0122, -0.0115, 0.0132], + [-0.0103, -0.0074, 0.0398, ..., -0.0136, -0.0275, 0.0045], + [-0.0069, 0.0102, -0.0359, ..., 0.0074, 0.0197, -0.0023]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.1.weight': tensor([[ 6.5918e-02, 6.3965e-02, 6.2988e-02, ..., 8.2970e-05, + 1.3161e-04, 8.1539e-05], + [-5.6396e-02, -5.4443e-02, -5.3711e-02, ..., -1.4591e-04, + -2.3365e-04, -1.6212e-04], + [-1.9409e-02, -1.8555e-02, -1.8433e-02, ..., 2.2173e-05, + 7.2956e-05, 8.2493e-05], + ..., + [ 6.8359e-02, 6.6406e-02, 6.4453e-02, ..., 6.8665e-05, + 1.0395e-04, 1.3733e-04], + [ 6.5430e-02, 6.3477e-02, 6.2256e-02, ..., 1.0109e-04, + -9.4771e-06, 6.9141e-05], + [ 7.7637e-02, 7.5684e-02, 7.4219e-02, ..., 3.5477e-04, + 3.3379e-04, 3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_self_attn.2.weight': tensor([[-0.0074, 0.0091, 0.0093, ..., -0.0056, -0.0071, -0.0049], + [-0.0017, 0.0016, 0.0010, ..., -0.0017, -0.0018, 0.0040], + [-0.0018, 0.0018, 0.0026, ..., -0.0016, -0.0017, -0.0002], + ..., + [-0.0120, 0.0197, 0.0259, ..., -0.0086, -0.0129, 0.0092], + [-0.0010, 0.0033, 0.0074, ..., -0.0016, -0.0031, -0.0171], + [ 0.0159, -0.0179, -0.0193, ..., 0.0148, 0.0153, 0.0121]], + dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.1.weight': tensor([[-1.5259e-02, -1.4465e-02, -1.3916e-02, ..., 6.4373e-05, + 9.8705e-05, 1.4591e-04], + [ 4.9316e-02, 4.8096e-02, 4.7119e-02, ..., -2.1935e-05, + -8.1539e-05, -5.0306e-05], + [ 2.6245e-02, 2.5879e-02, 2.5024e-02, ..., 1.8215e-04, + 1.0204e-04, 1.1015e-04], + ..., + [-6.6833e-03, -6.6223e-03, -6.3782e-03, ..., 2.2507e-04, + 2.3746e-04, 1.4305e-04], + [ 3.4027e-03, 3.3875e-03, 3.1586e-03, ..., -1.9073e-04, + -1.6022e-04, -2.6703e-04], + [-2.4719e-03, -2.1057e-03, -2.0752e-03, ..., -1.3161e-04, + -1.3447e-04, -1.8215e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_cross_attn.2.weight': tensor([[ 8.4400e-05, 7.9956e-03, -1.0803e-02, ..., -1.1292e-02, + 9.4604e-03, 1.1719e-02], + [-8.7738e-04, -5.8289e-03, 5.8594e-03, ..., 8.3618e-03, + -2.0905e-03, -7.1411e-03], + [ 2.9449e-03, 2.9755e-03, -3.4027e-03, ..., -2.1362e-04, + 2.2583e-03, 2.2583e-03], + ..., + [ 9.5215e-03, 2.3438e-02, 1.4954e-03, ..., -9.3994e-03, + 6.1035e-03, 2.9449e-03], + [ 1.6846e-02, 1.0376e-02, -1.1292e-02, ..., -1.1230e-02, + 4.8523e-03, 1.1108e-02], + [-1.7700e-02, -9.2773e-03, 9.9487e-03, ..., 1.5625e-02, + -8.3618e-03, -1.6602e-02]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.1.weight': tensor([[-5.2490e-03, -5.0964e-03, -4.9133e-03, ..., -4.2915e-05, + 8.2016e-05, -8.4400e-05], + [ 7.2754e-02, 7.1289e-02, 6.9336e-02, ..., -2.0504e-04, + -1.3924e-04, -2.2888e-04], + [ 3.6133e-02, 3.5400e-02, 3.4424e-02, ..., -3.1710e-05, + 2.4199e-05, -8.7261e-05], + ..., + [ 7.2754e-02, 7.1289e-02, 7.0312e-02, ..., -2.6321e-04, + -2.8038e-04, -2.7466e-04], + [ 2.9907e-02, 2.8809e-02, 2.8320e-02, ..., 3.4142e-04, + 3.1090e-04, 2.9945e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., -1.5831e-04, + -1.6212e-04, -2.2316e-04]], dtype=torch.bfloat16), 'net.blocks.7.adaln_modulation_mlp.2.weight': tensor([[ 7.3853e-03, 1.0376e-02, -1.1978e-03, ..., 1.4648e-02, + -1.7822e-02, 1.0803e-02], + [-1.3733e-03, 5.8365e-04, 1.3657e-03, ..., 5.4550e-04, + 6.1989e-06, 5.4932e-04], + [ 3.6926e-03, 6.7139e-04, -3.4485e-03, ..., 1.3885e-03, + 3.7842e-03, 8.5068e-04], + ..., + [ 1.2634e-02, -1.8311e-02, -1.3855e-02, ..., -2.1606e-02, + 7.2937e-03, -1.8677e-02], + [-1.1780e-02, 2.7344e-02, 1.7090e-02, ..., 3.2227e-02, + -6.9275e-03, 2.8564e-02], + [-8.3008e-03, 1.6846e-02, 1.2268e-02, ..., 1.9531e-02, + -2.2583e-03, 1.7700e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_proj.weight': tensor([[-0.0085, -0.0053, -0.0015, ..., 0.0052, -0.0004, 0.0079], + [-0.0032, -0.0175, 0.0178, ..., -0.0276, -0.0208, 0.0132], + [ 0.0352, 0.0175, -0.0134, ..., 0.0139, -0.0103, 0.0050], + ..., + [ 0.0022, -0.0299, -0.0077, ..., -0.0094, 0.0276, 0.0042], + [-0.0070, -0.0068, 0.0165, ..., 0.0175, -0.0311, -0.0188], + [ 0.0097, -0.0031, 0.0017, ..., 0.0065, -0.0126, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm.weight': tensor([1.3359, 0.6758, 1.6875, 1.0781, 0.3594, 0.6953, 0.2158, 0.6172, 1.1797, + 1.2734, 1.0547, 1.2891, 1.2812, 1.2578, 1.2344, 1.2422, 1.2812, 1.2344, + 1.2266, 1.2812, 1.2422, 1.1719, 2.1094, 1.8438, 0.4434, 0.3418, 0.1475, + 1.3750, 1.2812, 0.7109, 1.2734, 1.2031, 0.9961, 1.0078, 1.1719, 1.2734, + 1.1875, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.0938, 1.6250, + 1.5078, 1.0938, 0.6797, 1.4688, 1.1562, 0.6758, 1.2656, 1.2891, 1.2109, + 1.2969, 1.2266, 1.1797, 1.2188, 0.1172, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 1.4453, 1.1797, 0.6367, 1.3672, 0.9180, 0.3633, 0.7852, 1.1719, + 1.1641, 1.2266, 0.3340, 0.7852, 1.2109, 1.2812, 1.2656, 0.6797, 1.2266, + 1.0312, 1.2422, 1.2656, 1.0156, 1.2266, 1.0469, 1.2344, 0.8320, 0.6562, + 0.6875, 1.2109, 1.2891, 1.2656, 1.2188, 1.1094, 1.2344, 1.2188, 1.2109, + 1.1641, 1.1172, 1.1562, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.9375, + 0.8203, 1.5312, 1.6328, 0.4141, 1.2969, 1.1641, 1.3125, 1.3281, 1.1797, + 1.0859, 1.2578, 1.1719, 1.2578, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.k_proj.weight': tensor([[-0.0037, -0.0050, 0.0085, ..., -0.0046, -0.0146, 0.0031], + [-0.0143, 0.0094, -0.0223, ..., 0.0181, 0.0001, -0.0020], + [ 0.0500, -0.0013, -0.0109, ..., 0.0128, -0.0072, 0.0086], + ..., + [ 0.0065, -0.0029, 0.0005, ..., 0.0146, -0.0094, -0.0132], + [ 0.0123, 0.0031, -0.0056, ..., 0.0101, -0.0103, -0.0012], + [ 0.0159, 0.0103, 0.0167, ..., -0.0228, 0.0186, 0.0117]], + dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm.weight': tensor([1.0078, 0.3867, 1.2734, 0.7344, 0.6055, 0.7578, 0.1699, 0.6289, 1.2188, + 1.2578, 1.0547, 1.2891, 1.2891, 1.2578, 1.2344, 1.2422, 1.3047, 1.2266, + 1.2266, 1.2812, 1.2422, 1.1719, 1.5469, 1.6484, 0.6562, 0.5039, 0.2852, + 1.5078, 1.2812, 0.6602, 1.2812, 1.2109, 1.0000, 1.0156, 1.1797, 1.2891, + 1.1797, 1.2656, 1.1016, 1.2891, 1.1797, 1.2656, 1.2109, 1.3906, 1.2812, + 1.4062, 0.9727, 0.6836, 1.3594, 1.2734, 0.6484, 1.2969, 1.2734, 1.2188, + 1.2969, 1.2188, 1.1875, 1.2188, 0.1147, 1.2422, 1.0000, 0.9375, 1.2656, + 1.2344, 2.0625, 1.3438, 1.1406, 1.6484, 0.7266, 0.2637, 0.7930, 1.1719, + 1.1172, 1.2344, 0.3223, 0.7930, 1.2188, 1.2812, 1.2734, 0.6797, 1.2344, + 1.0391, 1.2422, 1.2734, 1.0078, 1.2266, 1.2969, 1.2578, 0.7891, 0.6445, + 0.6641, 1.4453, 1.2734, 1.2734, 1.2188, 1.1016, 1.2422, 1.2188, 1.2109, + 1.1641, 1.1094, 1.1484, 1.2422, 1.2656, 1.2188, 1.1875, 1.2734, 1.5000, + 0.9727, 1.3125, 1.7656, 0.4004, 1.2578, 1.2656, 1.3125, 1.3281, 1.1641, + 1.0625, 1.2578, 1.1719, 1.2500, 1.2891, 1.2422, 1.1719, 1.2266, 1.2188, + 1.1875, 1.1484], dtype=torch.bfloat16), 'net.blocks.8.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.self_attn.v_proj.weight': tensor([[-1.8997e-03, -9.8228e-05, -1.7334e-02, ..., -9.8877e-03, + 5.4626e-03, 1.5503e-02], + [ 2.1515e-03, 2.8809e-02, 7.6294e-04, ..., 4.2419e-03, + 1.8463e-03, 7.0496e-03], + [ 3.2043e-03, 4.3640e-03, -1.3672e-02, ..., -1.8188e-02, + 2.2827e-02, 1.0986e-02], + ..., + [ 5.7678e-03, -1.1353e-02, -9.7656e-03, ..., 3.2349e-03, + 3.0060e-03, 1.2817e-02], + [-5.7068e-03, -1.9287e-02, 1.2695e-02, ..., 9.3384e-03, + 1.4648e-03, -2.3438e-02], + [-1.5564e-02, -1.6357e-02, -4.9133e-03, ..., 7.4768e-03, + 2.3926e-02, 1.4221e-02]], dtype=torch.bfloat16), 'net.blocks.8.self_attn.output_proj.weight': tensor([[-0.0061, 0.0042, -0.0100, ..., -0.0044, -0.0087, -0.0037], + [-0.0019, -0.0113, -0.0066, ..., -0.0020, 0.0124, -0.0049], + [-0.0096, -0.0051, -0.0201, ..., 0.0039, 0.0055, -0.0171], + ..., + [-0.0242, -0.0183, -0.0111, ..., 0.0068, -0.0063, 0.0058], + [ 0.0022, 0.0121, 0.0172, ..., -0.0106, -0.0342, 0.0284], + [-0.0245, -0.0170, -0.0096, ..., 0.0104, 0.0081, -0.0046]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_proj.weight': tensor([[ 0.0132, 0.0173, 0.0242, ..., -0.0212, 0.0205, 0.0030], + [-0.0054, -0.0016, -0.0121, ..., 0.0189, 0.0144, 0.0003], + [-0.0069, -0.0035, -0.0017, ..., 0.0060, 0.0056, -0.0070], + ..., + [-0.0272, 0.0067, 0.0028, ..., -0.0058, -0.0126, 0.0056], + [ 0.0148, 0.0025, 0.0023, ..., -0.0129, -0.0107, -0.0178], + [-0.0006, -0.0116, 0.0050, ..., -0.0126, -0.0098, -0.0166]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm.weight': tensor([0.5938, 0.0859, 0.0972, 0.2695, 0.1040, 0.8672, 0.1846, 0.8398, 0.1289, + 0.0405, 0.8164, 0.1338, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0737, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8750, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2637, + 0.8555, 0.1006, 0.7227, 0.1719, 0.6719, 0.7656, 0.8164, 0.4941, 0.4668, + 0.0522, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7891, + 0.7266, 0.7969, 0.8711, 0.7539, 0.3027, 0.0615, 0.0874, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0388, 0.1543, 0.7852, 0.0957, 0.8711, 0.0574, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2383, 0.8555, 0.2656, 0.8398, 0.7695, + 0.2305, 0.0854, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0437, + 0.8359, 0.8398, 0.0850, 0.6367, 0.1045, 0.0439, 0.8203, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_proj.weight': tensor([[-0.0086, 0.0017, 0.0187, ..., 0.0064, -0.0110, -0.0036], + [ 0.0120, -0.0017, -0.0086, ..., -0.0031, -0.0104, -0.0162], + [ 0.0033, 0.0001, 0.0151, ..., 0.0033, 0.0076, 0.0018], + ..., + [ 0.0090, -0.0068, -0.0187, ..., 0.0162, 0.0122, -0.0037], + [ 0.0132, -0.0104, 0.0005, ..., 0.0203, 0.0099, -0.0009], + [-0.0015, 0.0035, -0.0047, ..., 0.0039, -0.0060, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm.weight': tensor([0.5938, 0.0859, 0.0977, 0.2695, 0.1035, 0.8672, 0.1846, 0.8398, 0.1270, + 0.0405, 0.8164, 0.1328, 0.0615, 0.6797, 0.0786, 0.3652, 0.7422, 0.8125, + 0.0742, 0.1152, 0.3633, 0.0635, 0.0977, 0.2246, 0.1689, 0.7500, 0.0879, + 0.7461, 0.6953, 0.1709, 0.8008, 0.0981, 0.3398, 0.2100, 0.3828, 0.8516, + 0.1787, 0.0535, 0.6523, 0.1660, 0.5234, 0.3867, 0.0420, 0.8047, 0.8711, + 0.7266, 0.5117, 0.2656, 0.8086, 0.5391, 0.6484, 0.0669, 0.7227, 0.1592, + 0.2500, 0.1128, 0.1377, 0.4824, 0.9102, 0.5625, 0.8125, 0.7656, 0.2617, + 0.8555, 0.0996, 0.7227, 0.1719, 0.6719, 0.7617, 0.8164, 0.4941, 0.4668, + 0.0515, 0.4824, 0.7656, 0.7266, 0.7695, 0.6719, 0.2676, 0.7852, 0.7852, + 0.7266, 0.7930, 0.8711, 0.7539, 0.3027, 0.0620, 0.0869, 0.1396, 0.2188, + 0.6016, 0.1299, 0.8672, 0.0393, 0.1543, 0.7812, 0.0957, 0.8711, 0.0576, + 0.6641, 0.6641, 0.0439, 0.1436, 0.2393, 0.8555, 0.2637, 0.8398, 0.7695, + 0.2295, 0.0859, 0.2500, 0.6172, 0.1904, 0.8398, 0.7266, 0.8281, 0.0439, + 0.8359, 0.8398, 0.0840, 0.6367, 0.1045, 0.0439, 0.8164, 0.7852, 0.5625, + 0.7930, 0.1670], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.v_proj.weight': tensor([[ 0.0079, 0.0080, -0.0094, ..., 0.0057, -0.0048, 0.0297], + [-0.0166, -0.0022, 0.0117, ..., 0.0120, 0.0320, 0.0231], + [-0.0017, -0.0045, -0.0015, ..., 0.0093, 0.0022, 0.0068], + ..., + [-0.0096, 0.0033, 0.0054, ..., -0.0068, -0.0080, 0.0120], + [ 0.0013, 0.0005, 0.0019, ..., -0.0064, 0.0016, -0.0061], + [-0.0026, -0.0051, 0.0095, ..., 0.0054, 0.0325, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.output_proj.weight': tensor([[ 0.0045, 0.0042, -0.0118, ..., -0.0188, 0.0069, -0.0039], + [ 0.0015, -0.0091, 0.0361, ..., -0.0084, -0.0036, 0.0100], + [ 0.0209, 0.0037, -0.0193, ..., -0.0160, 0.0133, 0.0114], + ..., + [-0.0094, -0.0006, 0.0125, ..., 0.0051, -0.0041, -0.0161], + [ 0.0028, -0.0101, -0.0079, ..., 0.0043, -0.0111, 0.0033], + [ 0.0127, 0.0021, -0.0182, ..., 0.0100, -0.0039, 0.0083]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img.weight': tensor([[-0.0046, -0.0081, 0.0317, ..., 0.0200, -0.0107, -0.0376], + [ 0.0356, 0.0378, 0.0135, ..., 0.0325, 0.0391, 0.0168], + [-0.0244, -0.0264, 0.0371, ..., 0.0027, 0.0050, -0.0322], + ..., + [ 0.0069, 0.0109, -0.0027, ..., 0.0366, -0.0206, 0.0374], + [-0.0054, -0.0080, 0.0038, ..., -0.0096, 0.0086, 0.0192], + [ 0.0099, -0.0178, -0.0038, ..., -0.0166, -0.0042, 0.0248]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.v_img.weight': tensor([[ 0.0063, -0.0096, -0.0344, ..., -0.0049, 0.0347, -0.0181], + [ 0.0262, -0.0366, -0.0093, ..., 0.0216, 0.0349, -0.0067], + [-0.0100, -0.0286, 0.0056, ..., 0.0061, -0.0282, -0.0006], + ..., + [-0.0128, -0.0016, 0.0043, ..., -0.0089, 0.0165, 0.0076], + [-0.0322, -0.0082, -0.0170, ..., 0.0195, 0.0212, -0.0135], + [-0.0139, 0.0070, -0.0170, ..., -0.0129, 0.0059, 0.0073]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img.weight': tensor([[ 0.0164, 0.0312, -0.0282, ..., -0.0130, 0.0201, 0.0703], + [ 0.0540, 0.0043, -0.0065, ..., -0.0483, 0.0374, 0.0097], + [-0.0576, 0.0115, 0.0374, ..., -0.0153, -0.0188, -0.0138], + ..., + [ 0.0262, 0.0003, -0.0069, ..., 0.0003, -0.0006, -0.0165], + [-0.0222, -0.0156, -0.0437, ..., -0.0065, -0.0165, -0.0282], + [-0.0130, -0.0146, 0.0029, ..., 0.0023, -0.0234, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.8.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.8.mlp.layer1.weight': tensor([[ 0.0039, 0.0114, 0.0150, ..., -0.0132, -0.0098, -0.0088], + [-0.0099, 0.0233, -0.0216, ..., 0.0064, -0.0011, -0.0088], + [-0.0088, -0.0040, -0.0075, ..., 0.0164, 0.0131, 0.0137], + ..., + [-0.0049, -0.0177, -0.0226, ..., 0.0153, 0.0200, 0.0116], + [ 0.0229, 0.0060, -0.0081, ..., -0.0055, 0.0050, 0.0137], + [ 0.0183, -0.0134, 0.0042, ..., -0.0025, 0.0030, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.8.mlp.layer2.weight': tensor([[ 0.0133, -0.0183, -0.0040, ..., -0.0037, 0.0330, 0.0162], + [ 0.0084, -0.0054, -0.0101, ..., 0.0051, -0.0197, 0.0064], + [ 0.0116, 0.0067, -0.0041, ..., -0.0227, 0.0327, 0.0145], + ..., + [-0.0044, 0.0054, 0.0204, ..., 0.0030, 0.0003, -0.0052], + [ 0.0146, 0.0081, 0.0059, ..., 0.0092, -0.0265, -0.0019], + [ 0.0013, 0.0058, -0.0168, ..., -0.0140, -0.0040, 0.0149]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.1.weight': tensor([[ 0.0557, 0.0544, 0.0525, ..., -0.0004, -0.0005, -0.0003], + [-0.0466, -0.0454, -0.0447, ..., 0.0004, 0.0005, 0.0003], + [ 0.0444, 0.0435, 0.0420, ..., -0.0004, -0.0005, -0.0003], + ..., + [-0.0664, -0.0654, -0.0635, ..., 0.0002, 0.0003, 0.0002], + [ 0.0864, 0.0840, 0.0820, ..., -0.0002, -0.0004, -0.0003], + [ 0.0664, 0.0654, 0.0640, ..., -0.0002, -0.0003, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_self_attn.2.weight': tensor([[-6.5613e-03, 8.5068e-04, -8.3160e-04, ..., 1.3733e-02, + -7.6599e-03, -1.3977e-02], + [ 5.2261e-04, -2.3460e-04, 3.2043e-04, ..., 3.9101e-05, + 3.3760e-04, 3.8624e-05], + [ 1.6556e-03, -1.5564e-03, 1.5488e-03, ..., -3.5706e-03, + 3.0060e-03, 3.2043e-03], + ..., + [-2.1851e-02, 2.3926e-02, -2.3926e-02, ..., 1.3611e-02, + -7.8125e-03, -1.4038e-02], + [-7.9956e-03, 9.6436e-03, -9.7656e-03, ..., 3.6926e-03, + -3.0396e-02, -3.6621e-03], + [ 2.9175e-02, -3.0640e-02, 3.0518e-02, ..., -2.3560e-02, + 1.2634e-02, 2.1606e-02]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.1.weight': tensor([[-3.0884e-02, -3.0029e-02, -2.8809e-02, ..., -2.0504e-04, + -2.0885e-04, -1.8311e-04], + [-3.5400e-02, -3.4912e-02, -3.3691e-02, ..., 2.2125e-04, + 1.7548e-04, 9.0122e-05], + [ 1.0559e-02, 1.0742e-02, 1.0864e-02, ..., -1.1301e-04, + -1.8311e-04, -3.0518e-04], + ..., + [-1.8433e-02, -1.7700e-02, -1.6846e-02, ..., 1.6594e-04, + 2.3937e-04, 1.9741e-04], + [ 7.5684e-02, 7.2754e-02, 7.0312e-02, ..., -2.0504e-04, + -1.1206e-04, -2.3746e-04], + [ 1.7456e-02, 1.6968e-02, 1.6846e-02, ..., 3.2806e-04, + 2.7275e-04, 3.2043e-04]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0013, 0.0054, 0.0030, ..., -0.0168, 0.0173, 0.0041], + [ 0.0005, 0.0011, -0.0003, ..., 0.0016, 0.0005, 0.0003], + [ 0.0067, 0.0087, 0.0012, ..., -0.0032, -0.0054, -0.0037], + ..., + [ 0.0118, -0.0201, 0.0134, ..., -0.0009, 0.0153, 0.0147], + [ 0.0172, 0.0023, 0.0186, ..., 0.0079, -0.0072, 0.0057], + [-0.0140, 0.0114, -0.0183, ..., -0.0090, 0.0034, -0.0110]], + dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.1.weight': tensor([[ 7.5684e-02, 7.3242e-02, 7.1777e-02, ..., -5.6458e-04, + -5.0735e-04, -5.8746e-04], + [ 3.2471e-02, 3.1250e-02, 3.0640e-02, ..., 3.2806e-04, + 3.6240e-04, 3.3760e-04], + [-6.6895e-02, -6.5430e-02, -6.4453e-02, ..., 2.0790e-04, + 2.6512e-04, 1.6689e-04], + ..., + [ 5.9814e-02, 5.8838e-02, 5.7129e-02, ..., -2.2697e-04, + -3.0136e-04, -2.2888e-04], + [ 6.0303e-02, 5.9326e-02, 5.8105e-02, ..., -1.4782e-04, + -1.4973e-04, -9.4414e-05], + [ 5.1758e-02, 5.0537e-02, 5.0049e-02, ..., -7.4863e-05, + -1.4782e-04, -7.3433e-05]], dtype=torch.bfloat16), 'net.blocks.8.adaln_modulation_mlp.2.weight': tensor([[ 3.5248e-03, 8.0466e-06, -2.3956e-03, ..., 1.5335e-03, + 1.4877e-03, 1.0834e-03], + [ 1.0529e-03, -2.4605e-04, -6.7520e-04, ..., 1.3542e-04, + 3.2425e-04, -7.1526e-06], + [ 5.1880e-03, 7.0190e-04, -4.3335e-03, ..., 3.9978e-03, + 4.2419e-03, 2.7313e-03], + ..., + [-2.1606e-02, 7.3853e-03, 2.0142e-02, ..., -1.8311e-02, + -1.7334e-02, -1.6235e-02], + [-2.6978e-02, 1.3962e-03, 2.0264e-02, ..., -1.7334e-02, + -1.7578e-02, -1.5564e-02], + [ 3.6865e-02, 1.0864e-02, -3.0029e-02, ..., 2.7710e-02, + 2.6733e-02, 2.4780e-02]], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_proj.weight': tensor([[ 0.0255, 0.0008, -0.0262, ..., 0.0005, 0.0199, -0.0038], + [ 0.0003, 0.0074, 0.0082, ..., -0.0078, 0.0027, 0.0182], + [ 0.0352, 0.0017, -0.0055, ..., -0.0036, -0.0016, 0.0087], + ..., + [ 0.0114, 0.0066, -0.0065, ..., 0.0108, -0.0193, 0.0034], + [-0.0093, -0.0281, -0.0063, ..., -0.0109, 0.0078, 0.0060], + [ 0.0112, -0.0034, 0.0018, ..., -0.0014, 0.0041, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm.weight': tensor([1.6250, 1.1875, 1.0312, 0.9297, 0.6680, 1.2812, 0.4941, 1.2578, 0.4512, + 1.1953, 1.1250, 1.1172, 0.6562, 1.0703, 0.8711, 1.1562, 1.0938, 1.0391, + 1.0859, 0.5781, 1.1719, 0.0986, 1.1328, 1.6797, 1.1172, 1.7344, 0.8359, + 0.1504, 1.1562, 1.0078, 0.4434, 0.9297, 1.0781, 0.5000, 1.1562, 0.9062, + 0.5312, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1475, 1.0625, 1.5781, + 0.8984, 1.3125, 0.6172, 1.6016, 1.0000, 0.1670, 0.1523, 0.0698, 0.8320, + 0.5469, 1.0859, 0.5859, 0.6172, 1.1172, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 0.8789, 1.4609, 1.4922, 1.4766, 0.2637, 0.9297, 0.3418, 0.5352, + 0.2715, 0.6797, 0.9883, 1.0156, 0.1172, 1.1875, 0.9961, 1.1094, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.7578, 1.1797, 0.7930, 0.8359, + 1.2812, 0.6016, 0.5234, 1.2969, 0.4668, 1.0547, 1.0781, 0.7344, 1.0781, + 0.9609, 1.1953, 1.0000, 1.0078, 0.1348, 1.1250, 0.9336, 0.0903, 1.4844, + 1.1406, 1.1094, 1.6250, 0.1348, 0.8477, 0.3574, 0.0967, 0.1309, 0.0527, + 1.1094, 0.8633, 0.2637, 1.0469, 0.8164, 0.9609, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.k_proj.weight': tensor([[ 0.0071, -0.0096, -0.0069, ..., -0.0236, 0.0145, 0.0022], + [ 0.0170, -0.0038, 0.0330, ..., 0.0112, -0.0102, 0.0020], + [ 0.0150, -0.0031, 0.0026, ..., 0.0364, 0.0256, 0.0038], + ..., + [ 0.0047, 0.0272, 0.0116, ..., 0.0040, 0.0060, -0.0048], + [-0.0074, -0.0253, -0.0109, ..., -0.0093, -0.0148, 0.0050], + [ 0.0008, 0.0037, -0.0030, ..., 0.0004, 0.0091, -0.0063]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm.weight': tensor([1.2188, 1.0000, 1.2734, 1.2188, 0.6602, 1.1875, 0.5312, 1.2656, 0.4492, + 1.1875, 1.1250, 1.1250, 0.6523, 1.0859, 0.8750, 1.1562, 1.0938, 1.0469, + 1.0859, 0.5742, 1.1641, 0.0986, 1.3594, 1.4609, 1.0156, 1.4141, 0.8984, + 0.4277, 1.1641, 1.1953, 0.4277, 0.9453, 1.0703, 0.5000, 1.1562, 0.8984, + 0.5273, 1.1406, 1.0156, 1.1250, 1.0938, 1.0547, 0.1465, 1.1641, 1.4141, + 0.8203, 1.3125, 0.6055, 1.4531, 1.0469, 0.1357, 0.1318, 0.0732, 0.8477, + 0.5391, 1.0859, 0.5859, 0.6211, 1.1250, 1.1641, 1.0234, 1.1562, 1.1562, + 1.1094, 1.3203, 1.6406, 1.3906, 1.1719, 0.3047, 1.0781, 0.2793, 0.5664, + 0.2734, 0.7148, 1.0000, 1.0078, 0.1040, 1.1797, 0.9961, 1.1016, 0.9648, + 0.7617, 1.0781, 1.1094, 1.0625, 0.9297, 1.2734, 1.2969, 0.9336, 1.1484, + 1.2344, 0.5234, 0.5625, 1.2266, 0.4648, 1.0469, 1.0781, 0.7344, 1.0781, + 0.9531, 1.1953, 1.0000, 1.0078, 0.1387, 1.1250, 0.9375, 0.0908, 1.4609, + 1.3281, 1.0703, 1.4531, 0.2256, 1.0391, 0.5938, 0.2080, 0.2109, 0.0496, + 1.1172, 0.8789, 0.2441, 1.0469, 0.8164, 0.9648, 1.0859, 0.7188, 1.0859, + 1.1797, 0.3242], dtype=torch.bfloat16), 'net.blocks.9.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.self_attn.v_proj.weight': tensor([[ 0.0138, 0.0106, 0.0021, ..., 0.0237, -0.0221, -0.0153], + [-0.0139, 0.0021, -0.0168, ..., 0.0128, 0.0074, -0.0132], + [ 0.0021, 0.0146, -0.0104, ..., -0.0140, -0.0266, -0.0077], + ..., + [ 0.0007, 0.0161, -0.0065, ..., 0.0093, -0.0055, -0.0024], + [-0.0023, 0.0121, 0.0229, ..., -0.0165, -0.0110, -0.0063], + [ 0.0102, 0.0061, 0.0189, ..., 0.0198, -0.0102, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.self_attn.output_proj.weight': tensor([[ 0.0251, -0.0334, -0.0134, ..., -0.0025, -0.0095, -0.0077], + [ 0.0139, -0.0108, -0.0162, ..., -0.0018, 0.0069, -0.0270], + [-0.0209, 0.0054, -0.0082, ..., 0.0125, 0.0123, 0.0078], + ..., + [ 0.0100, 0.0113, -0.0019, ..., -0.0004, -0.0097, 0.0010], + [-0.0118, -0.0067, 0.0064, ..., 0.0087, -0.0088, -0.0211], + [ 0.0188, -0.0028, 0.0081, ..., -0.0070, -0.0065, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_proj.weight': tensor([[-0.0344, 0.0420, -0.0193, ..., -0.0181, -0.0120, 0.0189], + [-0.0024, 0.0100, 0.0276, ..., 0.0065, 0.0150, 0.0132], + [-0.0110, 0.0081, 0.0017, ..., -0.0034, 0.0081, 0.0018], + ..., + [-0.0215, -0.0076, -0.0189, ..., -0.0034, 0.0042, 0.0071], + [-0.0031, -0.0132, -0.0168, ..., -0.0050, 0.0154, 0.0103], + [-0.0063, 0.0004, -0.0050, ..., 0.0021, 0.0005, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm.weight': tensor([0.9102, 0.8242, 0.8203, 0.0381, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1128, 0.6367, 0.8672, 0.9141, 0.1572, 0.3359, 0.4785, 0.1650, 0.0786, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1514, 0.0491, 0.1387, 0.8711, + 0.8906, 0.5156, 0.2656, 0.2188, 0.1641, 0.4062, 0.1245, 0.2773, 0.0674, + 0.1396, 0.8516, 0.0547, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1621, 0.2422, 0.8125, 0.0679, 0.8711, 0.1475, 0.7383, + 0.8594, 0.1855, 0.1494, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1953, 0.0913, 0.1934, 0.7031, 0.6367, 0.2031, 0.2598, + 0.1348, 0.1504, 0.8867, 0.0825, 0.6719, 0.8398, 0.3066, 0.3535, 0.0559, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8125, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3184, 0.1299, 0.4512, 0.0718, 0.2266, + 0.6367, 0.4746, 0.1079, 0.2256, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_proj.weight': tensor([[ 0.0247, 0.0035, -0.0108, ..., 0.0119, -0.0325, -0.0002], + [-0.0211, -0.0079, 0.0009, ..., -0.0066, -0.0320, -0.0162], + [ 0.0022, 0.0096, 0.0113, ..., -0.0064, -0.0017, -0.0217], + ..., + [ 0.0003, 0.0017, 0.0176, ..., 0.0100, -0.0110, 0.0043], + [ 0.0033, -0.0101, 0.0239, ..., -0.0242, -0.0172, 0.0031], + [-0.0021, 0.0043, -0.0016, ..., -0.0003, -0.0008, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm.weight': tensor([0.9141, 0.8242, 0.8203, 0.0386, 0.3672, 0.1641, 0.7109, 0.3359, 0.2754, + 0.1133, 0.6367, 0.8672, 0.9141, 0.1562, 0.3359, 0.4785, 0.1660, 0.0791, + 0.1094, 0.3574, 0.5859, 0.8633, 0.8633, 0.1504, 0.0491, 0.1387, 0.8711, + 0.8867, 0.5117, 0.2656, 0.2188, 0.1641, 0.4062, 0.1240, 0.2773, 0.0674, + 0.1396, 0.8477, 0.0552, 0.8477, 0.1768, 0.8438, 0.8633, 0.1069, 0.8789, + 0.2578, 0.1875, 0.1611, 0.2422, 0.8086, 0.0674, 0.8711, 0.1465, 0.7344, + 0.8594, 0.1855, 0.1484, 0.8086, 0.3984, 0.7656, 0.7812, 0.8984, 0.8555, + 0.1699, 0.6406, 0.5781, 0.5547, 0.2354, 0.8789, 0.8633, 0.1216, 0.7734, + 0.0459, 0.4141, 0.1963, 0.0913, 0.1934, 0.7031, 0.6367, 0.2012, 0.2617, + 0.1348, 0.1504, 0.8867, 0.0815, 0.6719, 0.8398, 0.3066, 0.3535, 0.0552, + 0.8438, 0.8516, 0.6172, 0.7891, 0.4121, 0.8008, 0.0728, 0.6836, 0.8711, + 0.5117, 0.1807, 0.5117, 0.6016, 0.8906, 0.9805, 0.3203, 0.8164, 0.8164, + 0.2383, 0.2012, 0.1406, 0.7148, 0.3203, 0.1309, 0.4512, 0.0718, 0.2256, + 0.6367, 0.4746, 0.1079, 0.2266, 0.8516, 0.2314, 0.8750, 0.7656, 0.6641, + 0.9141, 0.1553], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.v_proj.weight': tensor([[-0.0309, 0.0030, 0.0020, ..., 0.0016, 0.0022, -0.0140], + [ 0.0098, -0.0055, -0.0075, ..., -0.0068, -0.0201, -0.0072], + [-0.0078, 0.0044, 0.0160, ..., 0.0087, -0.0064, -0.0027], + ..., + [ 0.0120, 0.0038, -0.0005, ..., 0.0041, 0.0173, -0.0245], + [ 0.0264, -0.0118, -0.0082, ..., -0.0040, -0.0056, 0.0059], + [ 0.0153, -0.0017, -0.0019, ..., -0.0117, -0.0186, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.output_proj.weight': tensor([[ 0.0074, 0.0104, 0.0176, ..., 0.0037, 0.0121, 0.0020], + [-0.0061, 0.0128, 0.0221, ..., 0.0019, 0.0109, -0.0201], + [-0.0042, 0.0097, -0.0205, ..., 0.0031, 0.0304, -0.0112], + ..., + [ 0.0036, 0.0014, 0.0021, ..., -0.0091, -0.0186, 0.0040], + [-0.0107, -0.0161, -0.0150, ..., -0.0020, 0.0081, 0.0162], + [-0.0007, -0.0131, 0.0192, ..., 0.0043, 0.0084, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img.weight': tensor([[-0.0212, 0.0244, 0.0216, ..., -0.0144, -0.0649, 0.0102], + [ 0.0249, 0.0205, -0.0050, ..., -0.0064, -0.0091, -0.0103], + [-0.0220, 0.0038, 0.0051, ..., 0.0306, 0.0381, -0.0036], + ..., + [ 0.0200, -0.0042, 0.0170, ..., 0.0118, 0.0320, -0.0157], + [ 0.0349, 0.0121, 0.0197, ..., 0.0398, 0.0315, -0.0217], + [-0.0081, 0.0256, 0.0078, ..., 0.0205, -0.0070, -0.0100]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.v_img.weight': tensor([[ 0.0228, 0.0065, 0.0199, ..., -0.0063, 0.0276, -0.0312], + [ 0.0178, -0.0055, 0.0096, ..., 0.0113, -0.0398, 0.0293], + [ 0.0430, -0.0120, 0.0085, ..., -0.0029, -0.0067, 0.0060], + ..., + [ 0.0193, -0.0109, -0.0112, ..., -0.0049, 0.0330, 0.0143], + [-0.0122, -0.0542, -0.0205, ..., 0.0034, -0.0159, 0.0320], + [ 0.0303, -0.0039, -0.0425, ..., -0.0064, -0.0342, -0.0337]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img.weight': tensor([[-0.0275, -0.0067, -0.0315, ..., 0.0525, -0.0052, -0.0349], + [-0.0031, -0.0032, -0.0156, ..., 0.0055, -0.0240, 0.0305], + [-0.0256, 0.0006, 0.0060, ..., -0.0459, -0.0146, -0.0165], + ..., + [ 0.0053, 0.0231, 0.0209, ..., 0.0073, 0.0050, 0.0461], + [-0.0078, 0.0006, 0.0060, ..., -0.0217, 0.0347, -0.0021], + [ 0.0051, 0.0234, -0.0025, ..., 0.0261, -0.0041, 0.0103]], + dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.9.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.9.mlp.layer1.weight': tensor([[ 0.0184, -0.0075, -0.0048, ..., -0.0077, 0.0137, -0.0113], + [ 0.0012, 0.0091, 0.0175, ..., -0.0082, -0.0063, -0.0457], + [ 0.0159, -0.0121, 0.0038, ..., -0.0043, 0.0128, 0.0320], + ..., + [ 0.0190, -0.0043, -0.0104, ..., 0.0059, 0.0025, 0.0203], + [ 0.0021, 0.0070, -0.0024, ..., 0.0018, 0.0170, 0.0013], + [ 0.0034, -0.0103, -0.0014, ..., 0.0042, -0.0156, 0.0038]], + dtype=torch.bfloat16), 'net.blocks.9.mlp.layer2.weight': tensor([[-0.0109, -0.0082, 0.0069, ..., 0.0070, 0.0081, 0.0129], + [-0.0015, -0.0038, 0.0027, ..., -0.0278, -0.0027, -0.0201], + [-0.0041, 0.0176, 0.0162, ..., 0.0233, -0.0054, 0.0006], + ..., + [ 0.0077, 0.0144, -0.0069, ..., -0.0044, 0.0021, 0.0059], + [ 0.0079, 0.0074, 0.0270, ..., 0.0049, -0.0121, 0.0187], + [-0.0070, -0.0227, -0.0015, ..., -0.0090, 0.0221, 0.0054]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.1.weight': tensor([[-4.4189e-02, -4.2969e-02, -4.1992e-02, ..., 5.5313e-05, + 2.8312e-06, 5.7697e-05], + [-4.6143e-02, -4.5166e-02, -4.4434e-02, ..., -6.1989e-05, + -5.4121e-05, 5.1975e-05], + [ 2.4109e-03, 2.4414e-03, 2.3956e-03, ..., -9.6321e-05, + -3.1710e-05, -5.8413e-05], + ..., + [-4.4250e-03, -4.2419e-03, -4.2725e-03, ..., 1.0610e-05, + -6.5267e-06, -6.6757e-05], + [ 4.2480e-02, 4.1748e-02, 4.1016e-02, ..., -4.2915e-05, + -1.4591e-04, -1.0872e-04], + [ 2.6855e-02, 2.6123e-02, 2.4536e-02, ..., -1.4782e-04, + -8.9645e-05, -4.3869e-05]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_self_attn.2.weight': tensor([[ 0.0003, -0.0006, -0.0015, ..., 0.0027, -0.0003, -0.0038], + [-0.0018, -0.0013, -0.0064, ..., -0.0002, 0.0018, -0.0019], + [-0.0004, -0.0041, -0.0009, ..., 0.0023, 0.0004, -0.0018], + ..., + [ 0.0270, -0.0052, -0.0014, ..., -0.0205, -0.0260, 0.0376], + [ 0.0120, 0.0062, 0.0152, ..., -0.0036, -0.0121, 0.0171], + [-0.0315, -0.0173, 0.0037, ..., 0.0037, 0.0320, -0.0305]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.1963e-02, 1.1597e-02, ..., 1.5831e-04, + 1.9646e-04, 1.6308e-04], + [-4.7363e-02, -4.6143e-02, -4.4434e-02, ..., 1.1969e-04, + 1.1158e-04, 4.1962e-05], + [-7.7148e-02, -7.5195e-02, -7.3242e-02, ..., 3.5477e-04, + 2.0981e-04, 2.4796e-04], + ..., + [ 1.7456e-02, 1.6235e-02, 1.5625e-02, ..., 1.6880e-04, + 2.8610e-04, 1.8311e-04], + [-1.5564e-02, -1.5564e-02, -1.4893e-02, ..., -1.6975e-04, + -7.3910e-05, -1.8692e-04], + [-4.4922e-02, -4.3457e-02, -4.2725e-02, ..., 1.7166e-04, + 1.7071e-04, 1.0061e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_cross_attn.2.weight': tensor([[-0.0004, -0.0025, 0.0071, ..., -0.0016, 0.0117, -0.0075], + [ 0.0022, 0.0018, -0.0089, ..., 0.0027, -0.0023, 0.0016], + [-0.0074, -0.0011, 0.0081, ..., -0.0062, -0.0014, 0.0011], + ..., + [-0.0072, -0.0179, -0.0042, ..., -0.0084, 0.0084, -0.0189], + [-0.0205, -0.0140, -0.0101, ..., -0.0198, 0.0153, -0.0117], + [ 0.0303, 0.0148, -0.0083, ..., 0.0311, -0.0195, 0.0154]], + dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.1.weight': tensor([[ 7.7637e-02, 7.5684e-02, 7.3730e-02, ..., -6.1798e-04, + -5.6458e-04, -5.9891e-04], + [ 6.5918e-02, 6.4453e-02, 6.2988e-02, ..., -5.8174e-05, + -5.9605e-05, 3.3975e-06], + [ 1.1780e-02, 1.1780e-02, 1.0986e-02, ..., -3.4332e-04, + -3.2997e-04, -3.6621e-04], + ..., + [ 2.5513e-02, 2.5024e-02, 2.4536e-02, ..., 1.7285e-05, + 3.5763e-05, -5.0545e-05], + [ 5.8838e-02, 5.6641e-02, 5.5420e-02, ..., -4.3297e-04, + -2.2411e-04, -2.8038e-04], + [-7.5195e-02, -7.3730e-02, -7.1777e-02, ..., 4.3678e-04, + 3.4904e-04, 5.4550e-04]], dtype=torch.bfloat16), 'net.blocks.9.adaln_modulation_mlp.2.weight': tensor([[ 2.1515e-03, 1.8768e-03, -2.7313e-03, ..., 1.6022e-03, + 6.7749e-03, 1.0300e-04], + [-7.6294e-05, -1.2054e-03, -1.6785e-03, ..., 1.1587e-04, + 1.3428e-03, 1.2684e-04], + [-1.4038e-03, -4.0283e-03, 9.5749e-04, ..., 9.5749e-04, + -1.6403e-03, 5.1498e-04], + ..., + [-3.0029e-02, -2.1973e-02, -1.0315e-02, ..., -6.6223e-03, + 1.0071e-02, 2.3804e-02], + [-2.2339e-02, -1.7212e-02, -1.1047e-02, ..., 9.5825e-03, + 4.9744e-03, 1.7212e-02], + [ 3.7109e-02, 3.3203e-02, 1.6235e-02, ..., 7.3853e-03, + -1.0193e-02, -3.0518e-02]], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_proj.weight': tensor([[-0.0156, -0.0066, 0.0003, ..., -0.0095, 0.0097, -0.0125], + [-0.0049, -0.0065, 0.0022, ..., -0.0066, 0.0098, -0.0020], + [ 0.0065, 0.0039, 0.0004, ..., -0.0131, 0.0027, -0.0022], + ..., + [ 0.0031, -0.0027, 0.0076, ..., -0.0085, -0.0140, 0.0117], + [-0.0425, -0.0243, 0.0183, ..., -0.0295, 0.0055, -0.0256], + [-0.0231, -0.0063, -0.0216, ..., 0.0048, 0.0065, 0.0269]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm.weight': tensor([0.6797, 0.8672, 0.6602, 1.0781, 1.0781, 0.3965, 0.3477, 0.9805, 1.2188, + 1.2891, 1.2812, 1.2734, 1.3594, 1.3750, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1172, 1.3281, 1.3438, 1.3359, 1.2500, 1.2656, 1.0938, 1.1797, 0.9844, + 0.9453, 1.3516, 1.3750, 1.3672, 1.3203, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.2656, 0.8867, + 0.7227, 1.0234, 1.5625, 0.9922, 0.7695, 1.3828, 1.2891, 0.0586, 1.3359, + 1.3594, 1.3047, 1.3281, 1.3906, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8984, 0.9883, 1.2344, 1.4609, 0.9766, 0.3379, 0.3418, 0.9062, + 1.3125, 1.2891, 1.3516, 1.2031, 1.3672, 1.3516, 0.7812, 1.3359, 1.3047, + 1.3203, 1.3906, 1.3906, 1.3359, 1.3359, 0.6602, 1.5312, 1.1719, 1.5625, + 1.1797, 1.1406, 1.2344, 1.3516, 1.3438, 1.3438, 1.3125, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7969, + 1.4766, 1.2266, 0.6914, 1.0000, 0.5938, 0.1846, 1.2969, 1.3281, 1.3750, + 1.3984, 1.2188, 1.3828, 1.3359, 1.3359, 1.3281, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3438], dtype=torch.bfloat16), 'net.blocks.10.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.k_proj.weight': tensor([[-0.0031, 0.0004, 0.0023, ..., -0.0124, 0.0266, 0.0156], + [ 0.0081, -0.0013, 0.0065, ..., -0.0076, -0.0117, 0.0067], + [ 0.0157, -0.0248, -0.0142, ..., -0.0082, 0.0021, 0.0054], + ..., + [ 0.0030, 0.0021, -0.0099, ..., -0.0190, -0.0137, -0.0094], + [-0.0320, -0.0121, 0.0055, ..., -0.0171, -0.0284, -0.0173], + [-0.0308, 0.0077, 0.0025, ..., 0.0114, 0.0278, 0.0060]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm.weight': tensor([0.6172, 0.8203, 0.8203, 1.0234, 1.1484, 0.3789, 0.2734, 1.0078, 1.1641, + 1.2891, 1.2969, 1.2734, 1.3516, 1.3672, 1.3594, 1.3359, 0.9062, 1.3047, + 1.1250, 1.3359, 1.3438, 1.3281, 1.2344, 1.2188, 1.2188, 1.1953, 0.8516, + 1.0312, 1.3594, 1.3359, 1.3672, 1.3281, 1.3750, 1.3359, 1.3672, 1.3281, + 1.3516, 1.3594, 1.3281, 1.3594, 1.3750, 1.1719, 1.2266, 1.3672, 1.0312, + 0.8633, 1.0547, 1.5391, 0.9453, 0.7852, 1.3203, 1.3125, 0.0522, 1.3516, + 1.3516, 1.3047, 1.3281, 1.3828, 1.3203, 1.3203, 1.3828, 1.3516, 1.3281, + 1.3516, 0.8711, 0.9883, 1.0156, 1.4531, 0.8477, 0.3750, 0.3887, 0.8398, + 1.3359, 1.3047, 1.3359, 1.2109, 1.3672, 1.3516, 0.7852, 1.3359, 1.2969, + 1.3203, 1.3906, 1.3906, 1.3281, 1.3359, 0.9141, 1.4531, 1.1562, 1.4766, + 1.2734, 1.1719, 1.2344, 1.4531, 1.3359, 1.3516, 1.3047, 1.3750, 1.3125, + 1.3359, 1.2969, 1.3516, 0.3711, 1.3203, 1.3125, 1.3203, 1.2734, 0.7070, + 1.3281, 1.2891, 0.6875, 1.1484, 0.6797, 0.0505, 1.4688, 1.3203, 1.3672, + 1.3828, 1.2031, 1.3828, 1.3359, 1.3359, 1.3359, 1.3594, 1.3672, 1.3594, + 1.3203, 1.3359], dtype=torch.bfloat16), 'net.blocks.10.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.self_attn.v_proj.weight': tensor([[ 0.0167, -0.0023, -0.0117, ..., -0.0022, 0.0032, -0.0042], + [-0.0066, -0.0168, -0.0089, ..., 0.0092, 0.0033, 0.0006], + [-0.0009, 0.0069, 0.0183, ..., 0.0067, 0.0347, -0.0043], + ..., + [-0.0013, 0.0267, -0.0076, ..., -0.0070, 0.0099, 0.0206], + [ 0.0115, -0.0223, 0.0113, ..., 0.0040, 0.0248, -0.0098], + [ 0.0099, -0.0026, 0.0140, ..., 0.0001, 0.0009, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.10.self_attn.output_proj.weight': tensor([[ 0.0147, -0.0176, -0.0103, ..., 0.0245, -0.0111, -0.0281], + [ 0.0068, 0.0109, -0.0106, ..., -0.0093, 0.0220, 0.0118], + [ 0.0037, -0.0055, 0.0120, ..., 0.0216, 0.0159, -0.0273], + ..., + [ 0.0040, 0.0098, -0.0530, ..., 0.0068, 0.0108, 0.0036], + [ 0.0140, 0.0234, 0.0187, ..., -0.0240, -0.0078, -0.0114], + [ 0.0187, -0.0193, -0.0066, ..., 0.0137, 0.0369, -0.0109]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_proj.weight': tensor([[ 0.0090, 0.0083, 0.0106, ..., -0.0093, -0.0087, -0.0058], + [ 0.0012, 0.0089, -0.0020, ..., 0.0074, -0.0099, 0.0034], + [-0.0006, -0.0034, -0.0164, ..., -0.0048, 0.0018, 0.0015], + ..., + [-0.0168, 0.0004, 0.0220, ..., -0.0080, 0.0195, 0.0080], + [-0.0227, 0.0133, 0.0310, ..., 0.0195, 0.0098, 0.0074], + [-0.0079, 0.0015, 0.0272, ..., 0.0052, 0.0034, -0.0150]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm.weight': tensor([0.3516, 0.1787, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2246, 0.9336, 0.0771, 0.3008, 0.6562, 0.9766, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3145, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0728, 0.2871, 0.9336, 0.2236, 0.9414, 0.1001, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8828, 0.9258, 0.3242, 0.7578, 0.7148, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0588, 0.9219, 0.1279, 0.9648, + 0.1445, 0.8594, 0.4316, 0.2090, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0840, 0.7695, 0.1592, 0.0923, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5469, 0.3086, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1875, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2852, 0.8281, 0.1309, 0.9648, 0.9570, 0.1650, 0.8711, 0.0718, + 0.8477, 0.9648, 0.2188, 0.1108, 0.4824, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8906, 0.2461, + 0.1138, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0752, 0.9648, + 0.9414, 0.8008], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_proj.weight': tensor([[-0.0102, 0.0038, -0.0060, ..., -0.0005, 0.0069, -0.0031], + [-0.0055, 0.0020, -0.0099, ..., -0.0070, 0.0032, -0.0045], + [ 0.0051, 0.0029, 0.0052, ..., -0.0071, -0.0109, 0.0262], + ..., + [ 0.0205, 0.0136, 0.0032, ..., 0.0052, 0.0054, 0.0199], + [-0.0337, -0.0014, -0.0056, ..., -0.0159, -0.0144, -0.0078], + [ 0.0176, 0.0023, 0.0070, ..., -0.0056, 0.0016, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm.weight': tensor([0.3516, 0.1797, 0.9297, 0.9766, 0.8516, 0.3848, 1.0078, 0.1885, 0.2988, + 0.1621, 0.8320, 0.9531, 0.2256, 0.9336, 0.0767, 0.3008, 0.6523, 0.9727, + 0.9062, 0.8125, 1.0312, 0.1196, 0.0967, 0.3164, 0.1826, 0.9727, 0.3008, + 0.1875, 0.8008, 0.0723, 0.2871, 0.9336, 0.2217, 0.9414, 0.1006, 0.2207, + 0.7305, 0.9531, 0.9609, 0.8906, 0.8789, 0.9258, 0.3242, 0.7578, 0.7188, + 0.3613, 0.2734, 0.0684, 0.8242, 0.9766, 0.0598, 0.9219, 0.1289, 0.9648, + 0.1445, 0.8594, 0.4336, 0.2080, 0.4199, 0.9336, 0.9414, 0.1748, 0.9297, + 0.0845, 0.7695, 0.1592, 0.0933, 0.9219, 0.9609, 0.7109, 0.1689, 0.2295, + 0.5508, 0.3105, 0.7266, 0.2773, 0.9375, 0.6875, 0.8242, 0.1807, 0.8398, + 0.5625, 0.4941, 0.1709, 0.1895, 0.8828, 0.2695, 0.7344, 0.9023, 0.8398, + 0.8477, 0.2871, 0.8281, 0.1309, 0.9648, 0.9570, 0.1641, 0.8711, 0.0718, + 0.8477, 0.9609, 0.2188, 0.1128, 0.4805, 0.8750, 0.4414, 0.3223, 0.8867, + 0.9648, 0.9531, 0.9453, 0.1758, 0.1094, 0.2871, 0.1157, 0.8945, 0.2461, + 0.1147, 0.0693, 0.9180, 0.9570, 0.6523, 0.8984, 0.6445, 0.0757, 0.9648, + 0.9414, 0.8047], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.v_proj.weight': tensor([[ 0.0615, -0.0014, -0.0132, ..., 0.0142, 0.0177, 0.0171], + [ 0.0199, -0.0013, 0.0005, ..., 0.0060, -0.0095, 0.0254], + [-0.0108, 0.0061, 0.0146, ..., 0.0055, 0.0008, 0.0481], + ..., + [-0.0088, 0.0087, -0.0064, ..., -0.0106, -0.0060, -0.0231], + [ 0.0023, 0.0041, 0.0172, ..., 0.0023, 0.0178, 0.0146], + [-0.0029, 0.0025, -0.0154, ..., -0.0086, 0.0172, -0.0051]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.output_proj.weight': tensor([[-0.0181, -0.0177, -0.0199, ..., 0.0182, 0.0233, -0.0168], + [-0.0170, 0.0015, 0.0027, ..., 0.0045, 0.0256, 0.0043], + [-0.0033, 0.0024, -0.0232, ..., -0.0113, 0.0030, -0.0049], + ..., + [ 0.0022, 0.0021, -0.0172, ..., -0.0079, 0.0102, 0.0065], + [-0.0072, -0.0183, 0.0031, ..., 0.0125, -0.0085, 0.0106], + [-0.0215, -0.0095, -0.0132, ..., 0.0037, 0.0045, 0.0087]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img.weight': tensor([[ 0.0410, 0.0187, 0.0366, ..., 0.0320, 0.0037, 0.0006], + [-0.0183, -0.0211, -0.0049, ..., -0.0145, 0.0146, -0.0162], + [-0.0060, -0.0160, 0.0062, ..., -0.0203, 0.0267, 0.0221], + ..., + [ 0.0308, 0.0098, 0.0064, ..., 0.0171, -0.0042, -0.0064], + [-0.0142, 0.0260, -0.0083, ..., 0.0067, 0.0104, -0.0166], + [ 0.0095, -0.0142, -0.0040, ..., 0.0062, 0.0078, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.v_img.weight': tensor([[-0.0074, 0.0143, -0.0065, ..., -0.0278, 0.0240, -0.0118], + [ 0.0012, 0.0188, -0.0160, ..., -0.0317, -0.0069, 0.0182], + [ 0.0071, 0.0310, -0.0093, ..., -0.0226, 0.0023, 0.0334], + ..., + [-0.0092, 0.0186, -0.0053, ..., 0.0228, 0.0063, 0.0050], + [ 0.0271, 0.0123, 0.0354, ..., -0.0062, 0.0198, 0.0143], + [ 0.0282, 0.0171, 0.0317, ..., -0.0113, 0.0403, -0.0238]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img.weight': tensor([[-0.0110, -0.0265, -0.0115, ..., 0.0049, 0.0038, 0.0132], + [-0.0332, -0.0047, 0.0306, ..., 0.0347, -0.0361, 0.0085], + [-0.0044, -0.0190, -0.0063, ..., 0.0166, -0.0135, 0.0137], + ..., + [-0.0200, -0.0009, 0.0057, ..., -0.0339, 0.0361, 0.0006], + [-0.0082, -0.0250, 0.0151, ..., -0.0035, -0.0162, 0.0220], + [-0.0075, 0.0022, 0.0205, ..., -0.0065, -0.0083, 0.0435]], + dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.10.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.10.mlp.layer1.weight': tensor([[ 0.0040, 0.0157, -0.0132, ..., 0.0161, 0.0234, -0.0043], + [ 0.0056, 0.0198, 0.0369, ..., 0.0012, -0.0149, 0.0176], + [ 0.0127, 0.0094, 0.0146, ..., 0.0040, 0.0166, 0.0128], + ..., + [ 0.0112, 0.0222, -0.0104, ..., 0.0027, -0.0011, 0.0260], + [ 0.0277, -0.0188, -0.0173, ..., -0.0243, -0.0026, 0.0104], + [ 0.0081, -0.0099, 0.0286, ..., -0.0045, 0.0127, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.10.mlp.layer2.weight': tensor([[ 0.0140, 0.0011, 0.0020, ..., -0.0099, 0.0142, 0.0138], + [-0.0155, -0.0162, 0.0091, ..., -0.0116, 0.0110, -0.0051], + [ 0.0037, 0.0193, -0.0104, ..., 0.0071, -0.0199, 0.0140], + ..., + [ 0.0076, -0.0088, -0.0028, ..., 0.0132, 0.0047, 0.0103], + [-0.0031, 0.0025, -0.0114, ..., 0.0071, -0.0289, -0.0121], + [-0.0086, -0.0095, 0.0012, ..., -0.0042, -0.0027, 0.0091]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.1.weight': tensor([[ 0.0757, 0.0747, 0.0723, ..., -0.0002, -0.0002, -0.0002], + [ 0.0693, 0.0679, 0.0664, ..., -0.0002, -0.0002, -0.0002], + [-0.0562, -0.0552, -0.0537, ..., 0.0006, 0.0007, 0.0006], + ..., + [ 0.0737, 0.0718, 0.0703, ..., -0.0003, -0.0003, -0.0003], + [ 0.0698, 0.0684, 0.0669, ..., -0.0002, -0.0002, -0.0001], + [-0.0243, -0.0236, -0.0231, ..., 0.0003, 0.0003, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_self_attn.2.weight': tensor([[-6.6528e-03, -3.5095e-03, 4.4861e-03, ..., -5.4016e-03, + -3.4637e-03, 4.1199e-03], + [ 1.5974e-05, -6.8283e-04, 8.7357e-04, ..., 1.0252e-04, + -6.5994e-04, 3.7842e-03], + [-9.5367e-04, -1.4496e-03, 3.4332e-03, ..., -1.5488e-03, + -9.2697e-04, 7.7724e-05], + ..., + [-3.8757e-03, -1.0437e-02, -3.6133e-02, ..., -3.9368e-03, + -1.0498e-02, 1.0437e-02], + [-5.1575e-03, -9.8877e-03, -1.9287e-02, ..., -6.4392e-03, + -8.4229e-03, 1.0498e-02], + [ 2.3682e-02, 2.8564e-02, 3.3691e-02, ..., 2.4170e-02, + 2.8687e-02, -6.9275e-03]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.1.weight': tensor([[ 9.8267e-03, 1.0437e-02, 1.0071e-02, ..., -1.9455e-04, + -1.0395e-04, -1.5640e-04], + [ 3.0212e-03, 3.6469e-03, 3.5858e-03, ..., -4.6492e-05, + -1.5259e-04, -1.5545e-04], + [-9.1797e-02, -8.9844e-02, -8.7891e-02, ..., -1.6117e-04, + -1.7929e-04, -1.2589e-04], + ..., + [ 1.1902e-02, 1.1963e-02, 1.1902e-02, ..., -1.4877e-04, + -9.7752e-05, -1.9264e-04], + [-1.9455e-03, -1.8616e-03, -1.9226e-03, ..., -7.9632e-05, + -1.6594e-04, -1.1492e-04], + [ 1.0010e-02, 1.0193e-02, 9.7046e-03, ..., -5.4121e-05, + -2.4414e-04, -1.8120e-04]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0031, 0.0028, -0.0002, ..., 0.0032, 0.0027, 0.0031], + [ 0.0014, 0.0014, 0.0009, ..., 0.0014, 0.0014, 0.0014], + [-0.0029, -0.0028, 0.0014, ..., -0.0029, -0.0027, -0.0030], + ..., + [-0.0112, -0.0114, 0.0097, ..., -0.0113, -0.0117, -0.0114], + [ 0.0078, 0.0073, -0.0265, ..., 0.0079, 0.0073, 0.0077], + [-0.0111, -0.0108, 0.0332, ..., -0.0112, -0.0111, -0.0112]], + dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.1.weight': tensor([[ 6.4453e-02, 6.2988e-02, 6.1279e-02, ..., -1.0920e-04, + -5.7697e-05, -1.6332e-05], + [-1.5991e-02, -1.5625e-02, -1.5076e-02, ..., -3.1471e-04, + -2.6894e-04, -3.8338e-04], + [-3.3691e-02, -3.2959e-02, -3.1738e-02, ..., -2.0504e-04, + -2.6512e-04, -2.5749e-04], + ..., + [-6.8848e-02, -6.7383e-02, -6.5430e-02, ..., 4.7302e-04, + 4.1580e-04, 4.2343e-04], + [ 9.1797e-02, 9.0332e-02, 8.8379e-02, ..., 3.5858e-04, + 3.5286e-04, 3.2997e-04], + [-5.2734e-02, -5.1025e-02, -5.0293e-02, ..., -5.6028e-05, + 1.2577e-05, 6.6280e-05]], dtype=torch.bfloat16), 'net.blocks.10.adaln_modulation_mlp.2.weight': tensor([[-2.7466e-03, -8.6975e-04, 3.7909e-05, ..., 8.5449e-03, + 5.5847e-03, -6.5308e-03], + [ 3.9673e-03, 3.6621e-04, -3.4943e-03, ..., 3.3875e-03, + -2.9602e-03, -3.6812e-04], + [ 1.9989e-03, -3.4714e-04, -3.8528e-04, ..., -2.7924e-03, + 5.8289e-03, -7.2021e-03], + ..., + [ 2.1240e-02, 2.4567e-03, -2.3071e-02, ..., 2.5513e-02, + 6.3477e-03, -6.9580e-03], + [ 1.0132e-02, -3.1128e-03, -1.1597e-02, ..., 1.8188e-02, + -1.5259e-02, 2.2339e-02], + [-1.8311e-02, -3.3112e-03, 2.2217e-02, ..., -4.5898e-02, + -9.6893e-04, 9.8419e-04]], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_proj.weight': tensor([[-0.0272, 0.0081, -0.0024, ..., -0.0297, 0.0171, 0.0171], + [-0.0195, 0.0037, -0.0065, ..., -0.0111, 0.0024, -0.0002], + [-0.0271, -0.0138, -0.0186, ..., -0.0056, -0.0162, 0.0037], + ..., + [ 0.0173, -0.0075, 0.0029, ..., -0.0080, 0.0123, 0.0110], + [-0.0056, -0.0023, 0.0240, ..., 0.0097, -0.0053, -0.0021], + [ 0.0156, -0.0002, -0.0046, ..., 0.0030, 0.0116, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm.weight': tensor([1.3984, 0.3398, 1.4922, 0.6914, 0.6484, 0.4180, 0.3809, 1.0781, 1.2344, + 0.9570, 1.2109, 0.4355, 0.8633, 1.1719, 1.2031, 0.4180, 0.3164, 0.1875, + 1.1641, 1.1484, 1.1562, 1.1484, 1.7500, 1.1406, 1.3438, 1.6641, 1.0703, + 0.1982, 0.9844, 0.6641, 0.4980, 1.1875, 0.2559, 1.1094, 0.0222, 1.1719, + 1.2578, 1.1797, 0.3770, 1.1094, 0.4668, 0.1826, 1.1250, 1.4844, 1.8750, + 1.3984, 0.8359, 1.7422, 0.3262, 0.8398, 1.0156, 0.4219, 0.2432, 1.0156, + 1.2031, 1.1406, 1.2188, 1.0625, 0.3184, 0.4355, 0.1045, 1.1406, 0.1221, + 0.8359, 0.9062, 1.6875, 0.5820, 1.3438, 1.2422, 0.1934, 0.4668, 1.1406, + 0.0811, 1.3203, 1.2109, 0.7891, 0.7070, 0.2178, 0.4922, 1.1641, 1.0938, + 1.1172, 1.1016, 1.1094, 0.1260, 1.2266, 1.2031, 1.6875, 1.2969, 1.1641, + 0.8750, 0.3984, 1.2656, 0.3281, 0.7695, 0.2852, 1.1172, 1.1250, 0.0400, + 0.5703, 1.0391, 0.2188, 1.2656, 1.2422, 1.2031, 0.0952, 1.1641, 1.2891, + 0.9883, 1.0938, 1.6719, 1.1406, 0.8945, 0.6406, 1.2500, 1.2812, 1.2969, + 1.1484, 0.6797, 0.2812, 0.1572, 0.2734, 1.2031, 1.1250, 1.2344, 0.1079, + 1.0391, 0.3105], dtype=torch.bfloat16), 'net.blocks.11.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.k_proj.weight': tensor([[-0.0398, 0.0042, 0.0115, ..., 0.0098, 0.0142, -0.0179], + [ 0.0017, 0.0086, 0.0139, ..., 0.0265, -0.0198, -0.0060], + [-0.0115, -0.0091, -0.0109, ..., -0.0114, -0.0055, 0.0037], + ..., + [-0.0089, -0.0004, -0.0106, ..., -0.0016, -0.0032, 0.0009], + [-0.0156, -0.0192, -0.0022, ..., -0.0240, 0.0126, -0.0198], + [-0.0004, 0.0021, -0.0070, ..., 0.0056, 0.0020, 0.0039]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm.weight': tensor([ 1.4453, 1.0547, 1.5078, 0.5195, 0.8086, 0.4160, 0.2832, 1.0938, + 1.2344, 0.9648, 1.2031, 0.4023, 0.8633, 1.1719, 1.2031, 0.4219, + 0.3164, 0.1875, 1.1641, 1.1484, 1.1562, 1.1484, 1.4844, 1.2188, + 1.1016, 1.5859, 1.0781, 0.3242, 1.1875, 0.8906, 0.6641, 1.1875, + 0.2578, 1.1016, 0.0239, 1.1719, 1.2500, 1.1797, 0.3770, 1.1094, + 0.4668, 0.1807, 1.1250, 1.6328, 1.4219, 1.6016, 1.1797, 1.6719, + 0.4395, 0.8125, 1.0469, 0.3965, 0.3105, 1.0078, 1.2031, 1.1328, + 1.2188, 1.0625, 0.3184, 0.4414, 0.1040, 1.1406, 0.1226, 0.8359, + 0.8477, 1.4453, 0.4219, 1.3516, 1.1641, 0.2480, 0.5586, 1.1328, + -0.0091, 1.3203, 1.2109, 0.8047, 0.7070, 0.2168, 0.4922, 1.1641, + 1.0938, 1.1250, 1.1016, 1.1172, 0.1250, 1.2266, 1.2812, 1.5781, + 1.4141, 1.2188, 0.9648, 0.4609, 1.1641, 0.3320, 0.6719, 0.3047, + 1.1172, 1.1250, 0.0376, 0.5703, 1.0312, 0.2168, 1.2656, 1.2500, + 1.2031, 0.0962, 1.1719, 1.1719, 1.3203, 0.9375, 1.3984, 1.2188, + 0.8203, 0.7266, 1.3047, 1.2812, 1.2891, 1.1406, 0.6758, 0.2891, + 0.1523, 0.2695, 1.1953, 1.1250, 1.2344, 0.1079, 1.0391, 0.3105], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.self_attn.v_proj.weight': tensor([[ 0.0137, 0.0042, -0.0182, ..., 0.0051, 0.0093, -0.0239], + [ 0.0132, -0.0092, 0.0037, ..., -0.0064, -0.0023, 0.0043], + [ 0.0078, -0.0215, -0.0101, ..., -0.0234, 0.0162, -0.0009], + ..., + [ 0.0015, 0.0266, 0.0132, ..., 0.0099, -0.0026, -0.0269], + [ 0.0219, 0.0299, 0.0037, ..., 0.0091, -0.0272, 0.0040], + [-0.0095, 0.0068, -0.0130, ..., -0.0100, 0.0037, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.11.self_attn.output_proj.weight': tensor([[ 7.1716e-03, 1.0437e-02, 1.7944e-02, ..., 2.9907e-02, + 8.9111e-03, -7.1716e-04], + [-2.1515e-03, 3.3722e-03, -3.3417e-03, ..., -2.0264e-02, + -2.8381e-03, 8.3618e-03], + [-2.2217e-02, -1.0376e-02, -1.5320e-02, ..., -9.2163e-03, + -2.6855e-03, 1.1597e-02], + ..., + [ 2.2583e-02, -2.0630e-02, -1.0986e-02, ..., -1.9165e-02, + -6.1417e-04, -2.2827e-02], + [-7.5073e-03, 4.2114e-03, -8.0566e-03, ..., 9.2506e-05, + -1.6602e-02, -2.3071e-02], + [ 1.2329e-02, -1.0681e-02, -2.8229e-03, ..., 1.9531e-02, + -6.2561e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_proj.weight': tensor([[ 0.0171, -0.0038, 0.0003, ..., 0.0006, 0.0156, -0.0204], + [-0.0097, -0.0275, 0.0029, ..., 0.0048, 0.0134, 0.0065], + [-0.0104, 0.0038, -0.0041, ..., -0.0142, 0.0254, -0.0099], + ..., + [ 0.0085, 0.0045, -0.0045, ..., 0.0056, -0.0107, 0.0103], + [-0.0044, 0.0118, -0.0015, ..., -0.0064, -0.0075, -0.0056], + [-0.0045, -0.0029, -0.0030, ..., -0.0010, 0.0120, 0.0334]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0325, 0.4336, 0.0732, 0.3867, 0.0576, 0.8672, + 0.8477, 0.1553, 0.7930, 0.0664, 0.0796, 0.2500, 0.1245, 0.4121, 0.4492, + 0.2031, 0.0537, 0.8594, 0.1885, 0.5156, 0.5898, 0.3535, 0.2480, 0.8359, + 0.1328, 0.0303, 0.2246, 0.0693, 0.8711, 0.0850, 0.2100, 0.0498, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1338, 0.4141, 0.8828, 0.0457, 0.0505, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0552, 0.2832, 0.3223, 0.1187, 0.0381, + 0.8281, 0.4785, 0.0284, 0.0659, 0.3809, 0.0981, 0.0615, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0408, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3789, 0.0684, 0.0535, 0.1484, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0718, 0.1064, 0.3027, 0.1299, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0520, 0.0320, 0.3672, 0.8242, 0.1147, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9688, 0.3281, 0.6445, 0.4883, 0.0300, 0.8477, 0.1719, 0.1660, 0.7773, + 0.5820, 0.0388, 0.0579, 0.7461, 0.8008, 0.1055, 0.0586, 0.7891, 0.2832, + 0.0942, 0.2158, 0.0225, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_proj.weight': tensor([[-0.0041, 0.0068, 0.0020, ..., -0.0137, -0.0233, 0.0098], + [-0.0022, -0.0021, -0.0033, ..., 0.0036, -0.0093, -0.0053], + [ 0.0092, -0.0020, 0.0090, ..., -0.0063, -0.0023, -0.0023], + ..., + [-0.0090, 0.0032, -0.0016, ..., -0.0035, -0.0099, 0.0045], + [-0.0133, 0.0009, 0.0162, ..., -0.0210, -0.0120, -0.0077], + [ 0.0150, 0.0083, 0.0058, ..., -0.0010, -0.0162, -0.0033]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm.weight': tensor([0.8945, 0.0688, 0.6875, 0.0327, 0.4336, 0.0732, 0.3867, 0.0579, 0.8672, + 0.8477, 0.1543, 0.7930, 0.0669, 0.0796, 0.2490, 0.1235, 0.4121, 0.4492, + 0.2031, 0.0530, 0.8633, 0.1885, 0.5156, 0.5898, 0.3555, 0.2480, 0.8359, + 0.1328, 0.0305, 0.2227, 0.0693, 0.8711, 0.0845, 0.2100, 0.0503, 0.8516, + 0.2871, 0.5859, 0.7812, 0.0566, 0.1348, 0.4141, 0.8828, 0.0454, 0.0513, + 0.2012, 0.0332, 0.8242, 0.7695, 0.0554, 0.2852, 0.3242, 0.1187, 0.0376, + 0.8281, 0.4727, 0.0282, 0.0654, 0.3809, 0.0977, 0.0613, 0.6602, 0.1055, + 0.7539, 0.2275, 0.8750, 0.1553, 0.0488, 0.0415, 0.1235, 0.0645, 0.7891, + 0.6367, 0.3828, 0.0679, 0.0535, 0.1494, 0.8320, 0.0031, 0.7891, 0.7422, + 0.6562, 0.0713, 0.1064, 0.3027, 0.1289, 0.6562, 0.7461, 0.6875, 0.4141, + 0.0515, 0.0315, 0.3672, 0.8242, 0.1143, 0.6562, 0.4961, 0.7852, 0.4727, + 0.9648, 0.3281, 0.6445, 0.4883, 0.0303, 0.8477, 0.1719, 0.1670, 0.7773, + 0.5820, 0.0393, 0.0579, 0.7461, 0.8008, 0.1055, 0.0588, 0.7891, 0.2832, + 0.0942, 0.2129, 0.0223, 0.2930, 0.3516, 0.1050, 0.8555, 0.7461, 0.0684, + 0.7852, 0.7812], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.v_proj.weight': tensor([[-0.0032, -0.0036, 0.0199, ..., -0.0042, -0.0015, -0.0261], + [ 0.0223, 0.0048, -0.0125, ..., 0.0003, 0.0220, -0.0128], + [ 0.0119, 0.0023, 0.0165, ..., 0.0018, -0.0145, 0.0199], + ..., + [ 0.0021, -0.0032, -0.0212, ..., 0.0104, -0.0109, 0.0171], + [ 0.0093, -0.0065, 0.0110, ..., 0.0119, -0.0035, 0.0008], + [-0.0128, -0.0046, -0.0024, ..., 0.0082, -0.0250, -0.0139]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.output_proj.weight': tensor([[ 0.0161, -0.0085, -0.0008, ..., -0.0114, -0.0208, -0.0237], + [ 0.0085, -0.0059, -0.0161, ..., -0.0156, 0.0159, 0.0058], + [ 0.0021, 0.0192, -0.0024, ..., -0.0175, 0.0020, -0.0150], + ..., + [-0.0038, 0.0029, -0.0154, ..., -0.0108, 0.0125, 0.0132], + [ 0.0039, -0.0096, 0.0212, ..., -0.0100, -0.0267, -0.0140], + [-0.0023, -0.0160, 0.0039, ..., -0.0128, 0.0030, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img.weight': tensor([[-0.0066, 0.0240, -0.0217, ..., 0.0172, 0.0045, 0.0109], + [-0.0386, 0.0103, 0.0045, ..., -0.0055, 0.0322, 0.0087], + [-0.0427, -0.0052, -0.0042, ..., 0.0062, 0.0110, -0.0135], + ..., + [-0.0085, -0.0044, -0.0264, ..., -0.0586, -0.0630, -0.0129], + [ 0.0042, 0.0287, 0.0420, ..., -0.0311, -0.0089, -0.0289], + [-0.0106, -0.0610, 0.0162, ..., -0.0291, -0.0166, -0.0055]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.v_img.weight': tensor([[ 0.0122, -0.0251, 0.0211, ..., 0.0199, -0.0083, -0.0114], + [ 0.0112, -0.0101, -0.0327, ..., 0.0447, 0.0074, 0.0076], + [ 0.0053, -0.0175, 0.0070, ..., -0.0082, 0.0004, -0.0060], + ..., + [ 0.0295, -0.0042, -0.0103, ..., 0.0100, -0.0160, -0.0288], + [ 0.0085, 0.0253, -0.0177, ..., 0.0098, 0.0081, 0.0056], + [-0.0139, -0.0064, 0.0136, ..., -0.0128, 0.0277, 0.0086]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img.weight': tensor([[-0.0371, 0.0498, 0.0245, ..., -0.0061, -0.0276, -0.0271], + [ 0.0135, 0.0064, 0.0236, ..., 0.0085, 0.0182, -0.0251], + [ 0.0085, 0.0234, 0.0076, ..., 0.0464, 0.0066, 0.0309], + ..., + [-0.0123, -0.0175, 0.0327, ..., 0.0132, 0.0025, 0.0142], + [-0.0161, 0.0259, -0.0223, ..., -0.0317, 0.0237, -0.0181], + [ 0.0366, -0.0034, -0.0444, ..., 0.0058, -0.0283, 0.0029]], + dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.11.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.11.mlp.layer1.weight': tensor([[-0.0203, 0.0092, -0.0026, ..., -0.0027, 0.0317, -0.0059], + [ 0.0178, 0.0020, 0.0038, ..., 0.0277, 0.0030, 0.0192], + [-0.0175, 0.0026, -0.0135, ..., -0.0020, -0.0260, -0.0042], + ..., + [ 0.0084, 0.0289, 0.0161, ..., 0.0203, -0.0137, -0.0025], + [-0.0110, -0.0065, -0.0167, ..., -0.0118, -0.0383, -0.0334], + [ 0.0136, 0.0161, 0.0086, ..., 0.0041, -0.0052, -0.0111]], + dtype=torch.bfloat16), 'net.blocks.11.mlp.layer2.weight': tensor([[ 0.0008, 0.0015, 0.0143, ..., -0.0084, -0.0017, 0.0132], + [-0.0164, -0.0056, -0.0019, ..., 0.0405, -0.0124, -0.0123], + [-0.0109, 0.0212, 0.0073, ..., -0.0293, -0.0025, 0.0035], + ..., + [-0.0122, 0.0077, 0.0038, ..., -0.0125, -0.0039, -0.0087], + [ 0.0225, 0.0068, -0.0255, ..., 0.0564, -0.0179, -0.0194], + [ 0.0063, 0.0150, 0.0037, ..., -0.0527, -0.0140, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.1.weight': tensor([[ 0.0613, 0.0598, 0.0588, ..., 0.0001, 0.0001, 0.0002], + [ 0.0645, 0.0630, 0.0620, ..., -0.0006, -0.0005, -0.0006], + [ 0.0845, 0.0825, 0.0806, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0143, -0.0142, -0.0143, ..., -0.0002, -0.0003, -0.0002], + [-0.0052, -0.0053, -0.0051, ..., -0.0002, -0.0002, -0.0002], + [ 0.0273, 0.0256, 0.0243, ..., -0.0008, -0.0007, -0.0008]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_self_attn.2.weight': tensor([[ 0.0022, 0.0027, -0.0041, ..., -0.0022, -0.0008, 0.0003], + [-0.0001, -0.0005, 0.0002, ..., -0.0043, 0.0022, -0.0062], + [-0.0037, -0.0037, 0.0028, ..., -0.0019, -0.0029, -0.0046], + ..., + [ 0.0041, 0.0115, -0.0013, ..., 0.0287, -0.0001, 0.0327], + [-0.0135, -0.0121, -0.0087, ..., 0.0183, -0.0007, 0.0184], + [ 0.0361, 0.0349, -0.0068, ..., -0.0312, 0.0008, -0.0282]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.1.weight': tensor([[ 3.7109e-02, 3.6377e-02, 3.5645e-02, ..., 7.0095e-05, + 1.2589e-04, 2.0981e-04], + [-2.0264e-02, -1.9897e-02, -1.9653e-02, ..., -1.2398e-04, + -1.7071e-04, -1.2684e-04], + [ 6.0425e-03, 6.3782e-03, 6.8970e-03, ..., 2.1172e-04, + -8.1062e-06, 4.6015e-05], + ..., + [ 3.8574e-02, 3.7842e-02, 3.7354e-02, ..., 2.5177e-04, + 1.2016e-04, 1.6880e-04], + [-3.2471e-02, -3.2227e-02, -3.1494e-02, ..., -1.4591e-04, + -1.7834e-04, -2.2411e-04], + [-2.0630e-02, -2.0386e-02, -2.0874e-02, ..., -1.0347e-04, + -1.4186e-05, -1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0036, -0.0020, ..., -0.0063, 0.0063, 0.0020], + [-0.0059, 0.0019, -0.0023, ..., -0.0107, 0.0080, 0.0011], + [ 0.0023, -0.0024, 0.0052, ..., 0.0004, -0.0004, -0.0054], + ..., + [ 0.0173, -0.0082, 0.0068, ..., 0.0209, -0.0197, -0.0062], + [ 0.0087, -0.0177, 0.0183, ..., 0.0049, -0.0023, -0.0192], + [-0.0114, 0.0242, -0.0267, ..., -0.0109, 0.0087, 0.0265]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.1.weight': tensor([[-0.0713, -0.0693, -0.0684, ..., -0.0002, -0.0003, -0.0003], + [-0.0752, -0.0742, -0.0723, ..., -0.0002, -0.0003, -0.0002], + [-0.0757, -0.0742, -0.0718, ..., -0.0002, -0.0002, -0.0002], + ..., + [ 0.0356, 0.0349, 0.0337, ..., -0.0004, -0.0004, -0.0003], + [ 0.0265, 0.0255, 0.0247, ..., -0.0004, -0.0004, -0.0003], + [ 0.0688, 0.0664, 0.0659, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.blocks.11.adaln_modulation_mlp.2.weight': tensor([[ 0.0007, -0.0002, -0.0031, ..., 0.0038, 0.0035, -0.0007], + [ 0.0035, 0.0046, 0.0063, ..., 0.0015, 0.0012, -0.0044], + [ 0.0005, -0.0004, -0.0023, ..., 0.0033, 0.0033, -0.0002], + ..., + [ 0.0233, 0.0203, 0.0110, ..., 0.0304, 0.0302, -0.0239], + [ 0.0280, 0.0299, 0.0269, ..., 0.0125, 0.0135, -0.0281], + [-0.0444, -0.0479, -0.0442, ..., -0.0211, -0.0227, 0.0442]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_proj.weight': tensor([[-0.0064, 0.0044, 0.0036, ..., 0.0049, -0.0037, 0.0001], + [ 0.0061, 0.0019, -0.0014, ..., 0.0153, -0.0231, 0.0087], + [ 0.0369, 0.0057, -0.0030, ..., 0.0165, 0.0289, 0.0028], + ..., + [-0.0097, -0.0041, 0.0037, ..., 0.0168, 0.0315, 0.0053], + [-0.0103, 0.0083, 0.0092, ..., 0.0096, 0.0044, 0.0579], + [-0.0059, -0.0227, -0.0172, ..., 0.0009, 0.0167, -0.0427]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm.weight': tensor([0.7383, 1.1484, 1.0156, 1.5078, 1.2500, 0.8320, 0.6680, 0.4023, 1.2266, + 1.3594, 1.1016, 1.3984, 1.4453, 1.4062, 0.5078, 1.3047, 0.2393, 1.3359, + 1.3750, 1.2969, 1.3281, 1.3828, 1.2891, 1.4609, 1.1562, 1.8203, 0.7773, + 0.9961, 1.3203, 1.1406, 1.3438, 1.3828, 1.3594, 1.3281, 1.3750, 1.3906, + 1.1406, 1.3672, 1.3359, 1.3828, 0.4824, 1.3750, 1.2891, 0.8945, 1.1641, + 1.5391, 1.2891, 1.6875, 1.1250, 0.7500, 1.0156, 1.3359, 1.3984, 1.4062, + 1.2812, 1.1406, 1.3750, 1.2656, 1.3359, 1.3750, 1.3750, 1.2109, 1.2656, + 0.3066, 0.8047, 1.4609, 0.8398, 1.1953, 0.7734, 0.9883, 0.2119, 1.3281, + 0.1074, 0.1494, 1.2500, 1.3828, 1.2734, 1.4062, 0.2324, 1.4453, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3438, 0.8867, 1.2344, 1.3906, 1.6562, + 0.9375, 0.7266, 1.1875, 1.3359, 1.3438, 1.3672, 1.3828, 1.4297, 1.2422, + 0.2441, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3750, 1.4453, + 1.4375, 1.1719, 1.6719, 1.6172, 1.5078, 0.7891, 0.8164, 1.3203, 1.3594, + 0.9102, 1.1875, 1.2422, 0.3750, 1.3047, 1.3984, 1.3594, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.k_proj.weight': tensor([[-0.0240, -0.0090, -0.0190, ..., -0.0271, -0.0101, -0.0112], + [ 0.0200, 0.0082, 0.0003, ..., -0.0245, -0.0231, -0.0151], + [-0.0188, -0.0078, -0.0332, ..., -0.0142, -0.0398, -0.0033], + ..., + [-0.0146, 0.0070, -0.0029, ..., -0.0004, 0.0073, -0.0108], + [-0.0287, -0.0031, 0.0171, ..., 0.0059, 0.0020, 0.0164], + [ 0.0212, -0.0172, -0.0048, ..., -0.0041, 0.0197, -0.0065]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm.weight': tensor([0.8008, 0.9375, 0.9492, 1.4844, 1.1562, 0.9688, 0.6680, 0.3730, 1.2266, + 1.3594, 1.1016, 1.3906, 1.4453, 1.4062, 0.5078, 1.3047, 0.2412, 1.3359, + 1.3828, 1.2969, 1.3281, 1.3828, 1.4297, 1.6641, 1.4766, 1.4766, 0.8711, + 0.8945, 1.4062, 1.1328, 1.3516, 1.3750, 1.3516, 1.3281, 1.3750, 1.3828, + 1.1328, 1.3672, 1.3281, 1.3828, 0.4824, 1.3750, 1.2891, 1.1484, 1.3047, + 1.3594, 1.3828, 1.3672, 1.1328, 0.8438, 1.0781, 1.3359, 1.3828, 1.3984, + 1.2891, 1.1406, 1.3828, 1.2656, 1.3438, 1.3750, 1.3750, 1.2109, 1.2578, + 0.3066, 0.6641, 1.6094, 0.9883, 1.1484, 0.8555, 0.7852, 0.1099, 1.3281, + 0.0806, 0.1279, 1.2578, 1.3906, 1.2969, 1.4062, 0.2324, 1.4531, 1.3594, + 1.2344, 1.3750, 1.4141, 1.3906, 1.3516, 1.3828, 1.5625, 0.9727, 1.4297, + 0.8828, 0.8555, 1.1016, 1.3828, 1.3672, 1.3750, 1.3828, 1.4219, 1.2344, + 0.2461, 1.2656, 1.2266, 1.4062, 1.2422, 1.3203, 1.3984, 1.3828, 1.4062, + 1.4531, 1.3984, 1.6250, 1.3672, 1.4766, 0.6875, 0.6602, 1.3047, 1.3594, + 0.9102, 1.1797, 1.2422, 0.3691, 1.3047, 1.3984, 1.3516, 1.2969, 1.4375, + 1.0703, 1.3281], dtype=torch.bfloat16), 'net.blocks.12.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.self_attn.v_proj.weight': tensor([[ 0.0101, 0.0029, -0.0187, ..., -0.0079, 0.0050, 0.0120], + [-0.0231, 0.0122, -0.0054, ..., -0.0040, 0.0114, -0.0156], + [ 0.0010, -0.0280, -0.0148, ..., 0.0035, -0.0170, 0.0300], + ..., + [ 0.0019, -0.0143, 0.0045, ..., 0.0015, -0.0021, -0.0293], + [-0.0127, -0.0168, -0.0131, ..., 0.0037, -0.0325, -0.0064], + [ 0.0040, -0.0003, -0.0096, ..., -0.0122, -0.0148, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.12.self_attn.output_proj.weight': tensor([[-0.0312, 0.0009, 0.0052, ..., -0.0072, -0.0393, 0.0248], + [ 0.0009, -0.0068, 0.0101, ..., 0.0201, -0.0283, -0.0030], + [ 0.0002, -0.0098, -0.0093, ..., -0.0095, 0.0042, 0.0013], + ..., + [-0.0203, -0.0026, 0.0117, ..., 0.0034, 0.0193, 0.0127], + [ 0.0027, 0.0065, 0.0042, ..., -0.0072, -0.0123, -0.0045], + [-0.0143, 0.0025, 0.0165, ..., 0.0243, -0.0142, 0.0052]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_proj.weight': tensor([[-0.0087, 0.0021, 0.0072, ..., -0.0115, -0.0024, 0.0167], + [-0.0137, -0.0053, -0.0187, ..., 0.0042, 0.0139, 0.0018], + [ 0.0104, -0.0019, 0.0157, ..., -0.0148, -0.0098, 0.0130], + ..., + [ 0.0027, 0.0059, 0.0007, ..., 0.0074, -0.0148, -0.0248], + [ 0.0228, -0.0077, -0.0044, ..., 0.0106, 0.0066, -0.0143], + [ 0.0017, -0.0132, -0.0197, ..., 0.0024, 0.0108, 0.0005]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0830, 0.9453, 0.8945, 0.7461, 0.1934, 0.0703, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1943, 0.9141, + 0.1660, 0.1318, 0.0840, 0.6289, 0.5430, 0.8906, 0.7852, 0.9062, 0.1738, + 0.2598, 0.2793, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1348, + 0.0481, 0.9141, 0.1787, 0.6367, 0.3398, 0.1465, 0.0444, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2227, 0.1108, 0.0820, 0.8867, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1118, 0.8594, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9102, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2441, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1963, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0713, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0820, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1021, 0.1133, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_proj.weight': tensor([[-0.0023, -0.0086, -0.0261, ..., -0.0128, 0.0051, -0.0056], + [ 0.0021, 0.0038, -0.0070, ..., -0.0184, -0.0108, -0.0062], + [-0.0125, 0.0022, 0.0049, ..., -0.0098, -0.0187, 0.0115], + ..., + [ 0.0110, -0.0074, -0.0004, ..., -0.0117, -0.0303, 0.0225], + [ 0.0033, 0.0178, 0.0226, ..., 0.0145, 0.0115, 0.0093], + [-0.0015, -0.0048, 0.0010, ..., -0.0053, -0.0061, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm.weight': tensor([0.9844, 0.4180, 0.6016, 0.9258, 0.9375, 0.9766, 0.6719, 0.9492, 0.8320, + 0.0850, 0.9453, 0.8945, 0.7461, 0.1934, 0.0708, 1.0469, 0.1157, 0.9062, + 0.0962, 0.1777, 0.9219, 0.1299, 0.4121, 0.8398, 0.5078, 0.1953, 0.9180, + 0.1660, 0.1318, 0.0850, 0.6289, 0.5430, 0.8906, 0.7891, 0.9062, 0.1738, + 0.2578, 0.2773, 0.1201, 0.9453, 0.1738, 0.7031, 0.3574, 0.9531, 0.1338, + 0.0481, 0.9180, 0.1787, 0.6367, 0.3398, 0.1465, 0.0439, 0.1914, 0.2520, + 0.0464, 0.4473, 0.2246, 0.1104, 0.0825, 0.8828, 0.9023, 0.6797, 0.9727, + 0.8477, 0.9453, 0.9609, 0.1592, 0.1152, 0.1104, 0.8633, 1.0234, 0.9023, + 0.9492, 0.9258, 0.3945, 0.1309, 0.9062, 0.5273, 0.2070, 0.8594, 0.8398, + 0.8672, 0.9492, 0.9141, 0.7891, 0.0977, 0.9492, 0.3555, 0.5742, 0.0747, + 1.0625, 0.4316, 0.2461, 0.3359, 0.8906, 0.0908, 0.8633, 0.7656, 0.1943, + 0.9102, 0.8906, 0.7969, 0.8516, 0.9609, 0.8125, 0.0732, 0.2412, 1.0000, + 0.8008, 0.0771, 0.7422, 0.6367, 0.8906, 0.1934, 0.0615, 0.0815, 0.1079, + 0.9219, 0.9688, 0.8125, 0.1025, 0.1143, 0.9102, 0.8594, 0.6875, 0.9102, + 0.9961, 0.1631], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.v_proj.weight': tensor([[-0.0014, 0.0048, -0.0020, ..., 0.0035, 0.0098, -0.0143], + [ 0.0019, -0.0030, 0.0056, ..., -0.0034, -0.0107, -0.0125], + [-0.0014, -0.0089, -0.0100, ..., -0.0074, -0.0117, -0.0179], + ..., + [ 0.0152, 0.0035, -0.0025, ..., -0.0045, 0.0197, -0.0068], + [-0.0021, -0.0053, -0.0001, ..., 0.0065, 0.0126, -0.0011], + [ 0.0100, 0.0011, -0.0037, ..., -0.0125, 0.0244, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.output_proj.weight': tensor([[-0.0091, -0.0046, 0.0198, ..., 0.0010, -0.0075, -0.0173], + [ 0.0204, 0.0052, 0.0022, ..., -0.0337, 0.0310, -0.0282], + [-0.0184, 0.0012, 0.0071, ..., -0.0070, 0.0014, -0.0092], + ..., + [-0.0282, 0.0291, 0.0023, ..., -0.0040, 0.0019, -0.0146], + [ 0.0136, -0.0011, -0.0070, ..., -0.0038, -0.0240, 0.0042], + [ 0.0226, 0.0008, -0.0020, ..., -0.0205, -0.0171, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img.weight': tensor([[ 0.0067, -0.0104, 0.0148, ..., -0.0742, -0.0073, 0.0161], + [-0.0079, -0.0133, -0.0060, ..., 0.0056, 0.0049, 0.0239], + [ 0.0189, 0.0221, -0.0090, ..., 0.0049, 0.0408, 0.0049], + ..., + [ 0.0057, 0.0247, 0.0396, ..., 0.0057, -0.0304, -0.0183], + [ 0.0299, -0.0115, -0.0417, ..., -0.0049, -0.0376, 0.0041], + [-0.0210, 0.0168, -0.0003, ..., 0.0063, 0.0178, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.v_img.weight': tensor([[-0.0018, 0.0242, -0.0083, ..., -0.0245, 0.0064, -0.0058], + [ 0.0190, -0.0127, 0.0332, ..., -0.0146, 0.0176, 0.0408], + [-0.0330, -0.0258, -0.0079, ..., 0.0003, -0.0093, 0.0182], + ..., + [-0.0430, -0.0403, -0.0087, ..., 0.0135, -0.0141, -0.0055], + [-0.0332, -0.0157, 0.0217, ..., 0.0225, -0.0095, 0.0378], + [ 0.0364, 0.0486, -0.0117, ..., -0.0061, 0.0122, -0.0309]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img.weight': tensor([[-0.0243, 0.0132, 0.0122, ..., -0.0160, -0.0272, 0.0299], + [ 0.0095, -0.0095, -0.0159, ..., 0.0210, 0.0486, -0.0142], + [ 0.0227, -0.0332, 0.0076, ..., 0.0101, -0.0223, -0.0067], + ..., + [ 0.0192, 0.0292, -0.0026, ..., 0.0176, -0.0197, 0.0059], + [ 0.0265, 0.0063, 0.0067, ..., 0.0286, -0.0018, -0.0267], + [-0.0184, -0.0088, -0.0049, ..., 0.0381, 0.0192, -0.0217]], + dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.12.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.12.mlp.layer1.weight': tensor([[ 0.0151, 0.0093, -0.0131, ..., -0.0075, 0.0183, -0.0447], + [ 0.0005, -0.0042, 0.0165, ..., 0.0006, -0.0050, -0.0203], + [ 0.0034, 0.0344, 0.0059, ..., -0.0170, -0.0211, 0.0309], + ..., + [-0.0081, -0.0322, 0.0058, ..., 0.0016, 0.0115, 0.0115], + [-0.0071, 0.0040, -0.0067, ..., 0.0198, 0.0096, -0.0192], + [-0.0077, -0.0123, 0.0051, ..., -0.0018, 0.0015, 0.0015]], + dtype=torch.bfloat16), 'net.blocks.12.mlp.layer2.weight': tensor([[-0.0118, 0.0117, -0.0198, ..., 0.0040, -0.0033, -0.0261], + [ 0.0366, -0.0093, 0.0102, ..., 0.0223, 0.0067, 0.0016], + [ 0.0042, -0.0031, -0.0053, ..., -0.0120, -0.0160, 0.0221], + ..., + [ 0.0211, -0.0054, -0.0065, ..., 0.0156, 0.0064, -0.0087], + [-0.0171, -0.0038, -0.0112, ..., 0.0003, 0.0024, 0.0143], + [-0.0330, 0.0126, 0.0093, ..., 0.0132, 0.0283, 0.0035]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.1.weight': tensor([[-5.3467e-02, -5.2734e-02, -5.1758e-02, ..., 1.9169e-04, + 1.4210e-04, 1.3828e-04], + [-6.2866e-03, -6.4697e-03, -6.7749e-03, ..., 3.1853e-04, + 2.2507e-04, 2.1172e-04], + [ 1.0742e-01, 1.0400e-01, 1.0254e-01, ..., 1.2040e-05, + 5.1260e-05, 1.2779e-04], + ..., + [-1.0352e-01, -9.9121e-02, -9.7656e-02, ..., -1.2577e-05, + 5.3883e-05, -4.7207e-05], + [-6.0303e-02, -5.8838e-02, -5.7129e-02, ..., 2.1267e-04, + 1.7452e-04, 8.3923e-05], + [-1.3672e-02, -1.3977e-02, -1.3977e-02, ..., 3.1090e-04, + 2.3079e-04, 2.8992e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_self_attn.2.weight': tensor([[ 0.0002, -0.0043, -0.0008, ..., 0.0004, 0.0020, -0.0042], + [-0.0051, -0.0084, 0.0016, ..., -0.0021, -0.0028, -0.0079], + [-0.0029, -0.0027, 0.0035, ..., -0.0027, -0.0030, -0.0027], + ..., + [ 0.0159, 0.0250, 0.0052, ..., -0.0070, 0.0123, 0.0242], + [ 0.0233, 0.0229, -0.0231, ..., 0.0189, 0.0227, 0.0229], + [-0.0381, -0.0344, 0.0087, ..., -0.0069, -0.0388, -0.0354]], + dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.1.weight': tensor([[-1.4282e-02, -1.3672e-02, -1.3977e-02, ..., 6.2943e-05, + -2.6077e-06, -1.2684e-04], + [ 1.6846e-02, 1.7334e-02, 1.7334e-02, ..., -7.2956e-05, + 6.5804e-05, 4.4823e-05], + [ 1.5625e-02, 1.5564e-02, 1.5503e-02, ..., 1.0967e-05, + -9.0003e-06, -1.3256e-04], + ..., + [ 4.0771e-02, 3.9551e-02, 3.8330e-02, ..., -1.4877e-04, + 7.2718e-06, 5.5075e-05], + [-1.6602e-02, -1.6113e-02, -1.5991e-02, ..., 9.0599e-05, + 3.0041e-05, -6.2943e-05], + [-1.7212e-02, -1.7090e-02, -1.6968e-02, ..., 1.3053e-05, + -4.3154e-05, -3.4809e-05]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_cross_attn.2.weight': tensor([[-2.7895e-05, 2.4605e-04, 3.3951e-04, ..., 3.2425e-04, + 1.6117e-04, -3.6049e-04], + [-1.3428e-03, 1.2436e-03, 2.3346e-03, ..., 1.8005e-03, + -2.2430e-03, -1.1597e-03], + [ 1.5945e-03, -1.4954e-03, -1.6174e-03, ..., -1.8158e-03, + 2.2888e-03, 1.5259e-03], + ..., + [-1.3000e-02, 1.3184e-02, 1.3306e-02, ..., 1.4893e-02, + -1.2756e-02, -1.3184e-02], + [-8.1177e-03, 8.3618e-03, 8.6060e-03, ..., 1.0010e-02, + -8.1787e-03, -8.4839e-03], + [ 1.4893e-02, -1.5137e-02, -1.5259e-02, ..., -1.7090e-02, + 1.4893e-02, 1.5259e-02]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-02, 6.2988e-02, 6.1523e-02, ..., -3.2997e-04, + -3.3188e-04, -2.3651e-04], + [ 7.1289e-02, 6.9336e-02, 6.7871e-02, ..., -1.8311e-04, + -3.0136e-04, -1.9169e-04], + [ 3.4668e-02, 3.3203e-02, 3.1738e-02, ..., 1.4019e-04, + -8.4639e-06, 1.1206e-04], + ..., + [-2.1362e-02, -2.0874e-02, -2.0142e-02, ..., -2.6894e-04, + -4.2200e-05, -1.6212e-04], + [-1.1279e-01, -1.1084e-01, -1.0791e-01, ..., -1.7166e-04, + -1.1969e-04, -1.1015e-04], + [ 3.9795e-02, 3.8086e-02, 3.6621e-02, ..., 5.5790e-05, + 1.1015e-04, 1.1063e-04]], dtype=torch.bfloat16), 'net.blocks.12.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, 0.0017, 0.0029, ..., -0.0007, 0.0053, 0.0024], + [-0.0001, 0.0017, 0.0048, ..., -0.0025, -0.0030, 0.0043], + [-0.0011, -0.0006, 0.0035, ..., -0.0001, 0.0028, 0.0034], + ..., + [-0.0063, 0.0045, 0.0315, ..., -0.0170, -0.0033, 0.0310], + [-0.0211, -0.0171, 0.0223, ..., 0.0045, 0.0012, 0.0227], + [ 0.0452, 0.0454, -0.0273, ..., 0.0064, -0.0017, -0.0265]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_proj.weight': tensor([[-0.0674, 0.0093, 0.0072, ..., -0.0315, -0.0103, 0.0176], + [ 0.0222, 0.0119, -0.0082, ..., -0.0209, 0.0042, -0.0049], + [-0.0068, -0.0231, -0.0129, ..., -0.0215, -0.0063, -0.0189], + ..., + [ 0.0051, -0.0063, -0.0062, ..., 0.0069, -0.0022, -0.0248], + [ 0.0101, 0.0166, -0.0052, ..., -0.0045, -0.0063, 0.0077], + [-0.0068, -0.0145, 0.0154, ..., -0.0315, 0.0009, 0.0096]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm.weight': tensor([1.5547, 0.9531, 1.4922, 1.5938, 1.3281, 0.8555, 1.2109, 1.1016, 1.3047, + 1.3516, 1.3359, 1.3359, 1.3672, 1.3906, 1.3906, 1.3594, 1.2891, 1.3281, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4062, 1.2969, 1.2734, 1.3828, 1.0625, + 0.6992, 1.0078, 0.8984, 1.3594, 1.3281, 1.3359, 1.2422, 1.2891, 1.3203, + 1.3672, 1.3828, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.2344, 0.9648, + 1.5781, 1.6484, 1.5156, 0.6367, 1.1172, 0.8047, 1.0234, 1.3281, 1.3672, + 1.3438, 1.2578, 1.3281, 1.1172, 1.3672, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 0.9180, 1.4766, 0.7383, 0.8867, 0.8008, 0.8867, 0.4531, 1.2344, + 1.1172, 1.2656, 1.3516, 1.3516, 1.3750, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3281, 1.3594, 1.2969, 1.2656, 1.3984, 1.3359, 1.6953, + 0.8047, 0.5430, 0.4551, 1.1641, 0.8750, 1.3984, 1.3750, 1.3516, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 0.9023, + 1.5078, 1.1797, 1.0938, 1.3359, 1.1250, 0.7539, 0.9727, 1.2734, 1.3203, + 1.2344, 1.2578, 1.2734, 1.2734, 1.3828, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.k_proj.weight': tensor([[-0.0520, 0.0226, -0.0070, ..., -0.0022, -0.0036, -0.0134], + [ 0.0270, -0.0011, 0.0100, ..., -0.0267, -0.0006, -0.0067], + [ 0.0057, -0.0179, 0.0142, ..., -0.0420, -0.0080, -0.0106], + ..., + [ 0.0272, -0.0010, 0.0031, ..., 0.0060, -0.0127, -0.0160], + [-0.0104, 0.0131, -0.0153, ..., -0.0019, 0.0094, -0.0031], + [-0.0168, 0.0142, 0.0094, ..., -0.0378, -0.0107, 0.0074]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm.weight': tensor([1.4219, 1.1172, 1.2734, 1.5469, 1.1641, 0.6758, 1.2266, 1.0547, 1.3281, + 1.3594, 1.3516, 1.3516, 1.3672, 1.3828, 1.3906, 1.3594, 1.2891, 1.3359, + 1.2969, 1.3203, 1.3125, 0.6875, 1.4453, 1.2422, 1.5000, 1.3359, 1.2188, + 1.0078, 1.1875, 1.3359, 1.3203, 1.3359, 1.3281, 1.2578, 1.2891, 1.3203, + 1.3672, 1.3750, 1.1250, 1.3359, 1.3359, 1.1562, 1.3438, 1.4141, 1.2188, + 1.1328, 1.5234, 1.5547, 0.8828, 1.1406, 1.0078, 1.5547, 1.3047, 1.3672, + 1.3359, 1.2578, 1.3203, 1.1172, 1.3594, 1.3594, 1.3281, 1.3672, 1.2266, + 0.2432, 1.1562, 1.4766, 1.0781, 0.8945, 0.8672, 0.9453, 0.3516, 1.2344, + 1.0859, 1.2578, 1.3438, 1.3438, 1.3672, 1.2734, 1.3203, 1.2734, 1.3047, + 1.3359, 1.2500, 1.3203, 1.3594, 1.2969, 1.2031, 1.5234, 1.1875, 1.6172, + 0.9492, 0.7891, 0.8125, 1.3594, 1.1406, 1.3906, 1.3750, 1.3438, 1.3750, + 1.3672, 1.4141, 1.3203, 1.3516, 1.3672, 1.2812, 1.3281, 1.3750, 1.1328, + 1.4375, 1.3281, 1.2578, 1.1641, 1.0469, 1.0938, 1.0859, 1.3125, 1.3203, + 1.2422, 1.2578, 1.2812, 1.2734, 1.3750, 1.0625, 1.3516, 1.2422, 1.2422, + 1.3359, 1.3750], dtype=torch.bfloat16), 'net.blocks.13.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.self_attn.v_proj.weight': tensor([[ 0.0027, 0.0041, -0.0002, ..., -0.0156, 0.0025, -0.0081], + [-0.0007, 0.0149, 0.0209, ..., 0.0132, 0.0142, -0.0123], + [ 0.0084, -0.0302, -0.0150, ..., 0.0046, -0.0011, 0.0063], + ..., + [ 0.0181, -0.0092, 0.0156, ..., -0.0084, 0.0176, 0.0075], + [ 0.0030, 0.0077, 0.0237, ..., 0.0096, -0.0029, 0.0142], + [-0.0110, 0.0042, -0.0153, ..., 0.0042, -0.0006, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.13.self_attn.output_proj.weight': tensor([[-0.0066, 0.0069, -0.0187, ..., -0.0039, 0.0014, -0.0272], + [-0.0060, -0.0016, 0.0125, ..., 0.0126, -0.0199, 0.0231], + [-0.0327, 0.0275, -0.0349, ..., 0.0131, -0.0251, -0.0142], + ..., + [ 0.0109, -0.0036, 0.0013, ..., -0.0227, -0.0107, 0.0104], + [ 0.0040, 0.0255, 0.0089, ..., -0.0069, 0.0095, -0.0042], + [-0.0044, 0.0232, 0.0042, ..., -0.0061, -0.0016, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_proj.weight': tensor([[-0.0146, 0.0012, 0.0236, ..., 0.0125, 0.0249, 0.0012], + [-0.0269, -0.0018, -0.0050, ..., 0.0075, 0.0208, 0.0225], + [-0.0052, -0.0059, -0.0237, ..., -0.0072, -0.0162, 0.0079], + ..., + [-0.0129, -0.0211, -0.0217, ..., 0.0106, 0.0267, -0.0020], + [ 0.0190, -0.0126, 0.0036, ..., 0.0069, -0.0135, -0.0095], + [-0.0081, -0.0110, 0.0049, ..., -0.0028, 0.0164, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm.weight': tensor([0.9727, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9062, 1.0391, + 0.6211, 0.1768, 1.0391, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2207, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1270, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0625, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9453, 1.0078, 0.2969, 0.1650, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0884, 1.0859, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2314, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0078, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_proj.weight': tensor([[ 0.0079, 0.0106, 0.0156, ..., -0.0049, -0.0126, 0.0181], + [-0.0142, -0.0033, -0.0146, ..., 0.0024, -0.0211, -0.0171], + [-0.0136, 0.0001, 0.0013, ..., -0.0081, 0.0070, -0.0070], + ..., + [ 0.0045, 0.0013, -0.0193, ..., -0.0047, 0.0161, 0.0009], + [ 0.0198, 0.0011, 0.0131, ..., -0.0095, 0.0050, -0.0026], + [-0.0217, -0.0055, -0.0162, ..., -0.0223, 0.0179, 0.0112]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm.weight': tensor([0.9766, 0.9414, 0.2930, 0.7305, 1.0156, 0.9297, 0.1689, 1.0391, 1.0234, + 0.8359, 0.8906, 1.0312, 0.3555, 0.9688, 0.4844, 1.0234, 0.9102, 1.0391, + 0.6211, 0.1768, 1.0469, 0.8555, 1.0312, 0.9492, 0.9805, 1.0625, 0.1963, + 0.9414, 0.9570, 0.2217, 1.0156, 0.2119, 0.9648, 0.1670, 1.0469, 1.0234, + 1.0234, 0.9727, 0.7461, 0.1875, 0.4434, 1.1016, 0.1260, 0.2598, 0.2773, + 0.2480, 0.9961, 0.9648, 0.7188, 1.0156, 1.0156, 0.9844, 0.9609, 1.0078, + 0.2422, 1.0859, 0.1206, 0.8711, 0.8047, 1.0312, 0.1748, 1.0312, 0.9570, + 1.0312, 1.0000, 0.6016, 0.2461, 0.9062, 1.0312, 1.0156, 1.0000, 1.0781, + 0.7852, 0.9922, 0.2617, 0.6094, 1.0703, 0.2178, 0.7031, 0.8906, 1.0859, + 0.4062, 0.9844, 0.9609, 1.0547, 0.4648, 0.8867, 1.0703, 0.5156, 0.2793, + 0.9219, 0.3516, 0.9492, 1.0078, 0.2969, 0.1641, 0.8398, 1.0078, 0.5117, + 0.8672, 0.2734, 1.0156, 0.9961, 1.0938, 0.4707, 0.5469, 0.9453, 1.0156, + 1.0625, 1.0000, 1.0078, 0.0879, 1.0938, 0.1650, 1.0391, 0.6641, 0.7031, + 1.0156, 0.9805, 0.2295, 1.0625, 0.0864, 0.8672, 0.2734, 0.7344, 1.0234, + 1.0781, 0.6914], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.v_proj.weight': tensor([[-0.0188, 0.0068, -0.0041, ..., -0.0038, 0.0014, -0.0156], + [-0.0098, -0.0053, -0.0164, ..., -0.0021, -0.0057, -0.0134], + [-0.0056, -0.0013, 0.0076, ..., -0.0162, 0.0173, -0.0132], + ..., + [ 0.0356, 0.0064, -0.0344, ..., -0.0078, 0.0297, 0.0115], + [-0.0223, 0.0014, 0.0070, ..., -0.0037, -0.0320, 0.0042], + [ 0.0099, 0.0053, -0.0017, ..., -0.0060, -0.0096, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.output_proj.weight': tensor([[ 0.0104, -0.0135, 0.0069, ..., -0.0076, 0.0126, 0.0009], + [-0.0038, 0.0073, -0.0010, ..., 0.0182, -0.0027, -0.0250], + [ 0.0035, -0.0025, -0.0157, ..., -0.0122, -0.0135, 0.0039], + ..., + [-0.0275, 0.0233, 0.0108, ..., -0.0094, 0.0056, -0.0048], + [ 0.0017, 0.0269, 0.0258, ..., 0.0166, 0.0025, -0.0101], + [-0.0265, 0.0164, -0.0106, ..., -0.0069, 0.0242, 0.0289]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img.weight': tensor([[-0.0083, -0.0159, 0.0084, ..., 0.0198, 0.0059, 0.0277], + [ 0.0172, 0.0039, 0.0039, ..., -0.0035, 0.0021, 0.0022], + [-0.0059, 0.0184, 0.0111, ..., 0.0099, -0.0046, -0.0077], + ..., + [-0.0050, -0.0112, -0.0041, ..., -0.0096, -0.0050, 0.0118], + [-0.0427, -0.0391, -0.0192, ..., 0.0089, -0.0238, -0.0084], + [-0.0042, -0.0106, -0.0222, ..., -0.0164, 0.0342, 0.0090]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.v_img.weight': tensor([[ 0.0216, -0.0226, -0.0349, ..., -0.0087, -0.0161, 0.0119], + [ 0.0041, 0.0074, 0.0312, ..., 0.0074, 0.0286, 0.0317], + [ 0.0214, -0.0145, 0.0229, ..., -0.0082, 0.0056, 0.0277], + ..., + [ 0.0170, -0.0082, 0.0471, ..., 0.0072, 0.0076, -0.0046], + [ 0.0330, 0.0391, 0.0254, ..., -0.0234, -0.0067, -0.0203], + [ 0.0079, 0.0083, -0.0037, ..., -0.0089, -0.0074, 0.0084]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img.weight': tensor([[ 0.0036, -0.0154, 0.0284, ..., -0.0104, 0.0135, -0.0042], + [ 0.0153, 0.0332, 0.0337, ..., 0.0223, -0.0165, 0.0051], + [-0.0068, 0.0215, -0.0315, ..., -0.0122, 0.0275, 0.0273], + ..., + [ 0.0069, 0.0077, 0.0280, ..., -0.0041, -0.0078, 0.0140], + [-0.0066, 0.0352, -0.0240, ..., -0.0281, 0.0294, -0.0208], + [-0.0206, 0.0159, -0.0092, ..., 0.0064, 0.0167, -0.0086]], + dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.13.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.13.mlp.layer1.weight': tensor([[ 0.0075, -0.0283, 0.0249, ..., -0.0120, 0.0206, 0.0237], + [ 0.0256, -0.0147, 0.0056, ..., -0.0054, -0.0114, -0.0049], + [-0.0061, 0.0105, 0.0134, ..., 0.0079, -0.0019, 0.0032], + ..., + [ 0.0067, -0.0112, -0.0056, ..., 0.0038, -0.0022, 0.0043], + [ 0.0026, -0.0234, -0.0126, ..., 0.0022, 0.0020, -0.0226], + [-0.0300, -0.0070, 0.0067, ..., -0.0094, -0.0093, -0.0029]], + dtype=torch.bfloat16), 'net.blocks.13.mlp.layer2.weight': tensor([[-4.5776e-05, -3.4027e-03, -3.7354e-02, ..., 4.8523e-03, + -1.7334e-02, -9.5215e-03], + [ 3.5645e-02, -1.6403e-03, -1.0925e-02, ..., 7.8125e-03, + 2.2583e-02, -1.9043e-02], + [-2.1362e-02, -7.5531e-04, 8.1787e-03, ..., 2.5330e-03, + -1.0254e-02, 2.3438e-02], + ..., + [-8.8501e-03, -1.8463e-03, 1.2329e-02, ..., -1.9165e-02, + 2.9449e-03, -8.4839e-03], + [ 2.2949e-02, -2.8229e-03, 5.7983e-03, ..., -2.3956e-03, + -7.8735e-03, -4.9561e-02], + [-2.4587e-06, 2.6855e-03, -1.0071e-02, ..., 1.7456e-02, + -1.8433e-02, 5.1270e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.1.weight': tensor([[ 1.7456e-02, 1.7700e-02, 1.7700e-02, ..., -1.3351e-04, + -1.3256e-04, -1.3828e-04], + [ 6.4941e-02, 6.2012e-02, 6.0303e-02, ..., -5.1498e-04, + -4.7302e-04, -5.1880e-04], + [ 6.2256e-02, 6.0547e-02, 6.0059e-02, ..., -1.7357e-04, + -3.2043e-04, -2.9182e-04], + ..., + [ 4.0283e-02, 4.0039e-02, 3.9062e-02, ..., -1.5831e-04, + -2.0123e-04, -1.9169e-04], + [-5.9814e-02, -5.8838e-02, -5.7861e-02, ..., 2.6131e-04, + 2.3651e-04, 2.9564e-04], + [-9.2773e-02, -9.1309e-02, -8.9355e-02, ..., -1.3638e-04, + -2.4986e-04, -9.5367e-05]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_self_attn.2.weight': tensor([[ 1.8768e-03, -1.8082e-03, 8.1253e-04, ..., 1.8463e-03, + -1.8311e-03, -4.5471e-03], + [ 2.4261e-03, -4.7607e-03, 2.6093e-03, ..., 2.6703e-03, + -2.5635e-03, 4.5776e-03], + [ 3.7842e-03, -9.0332e-03, 7.1049e-05, ..., 3.0365e-03, + -6.1798e-04, -9.7046e-03], + ..., + [ 1.6602e-02, 4.6082e-03, 3.1982e-02, ..., 2.2705e-02, + -3.0640e-02, -1.5411e-03], + [-2.3682e-02, 2.4902e-02, -1.8677e-02, ..., -2.2095e-02, + 1.9897e-02, 6.1951e-03], + [ 3.3203e-02, -2.6978e-02, 3.2959e-02, ..., 3.2471e-02, + -3.3691e-02, 1.0223e-03]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.1.weight': tensor([[ 7.2021e-03, 7.0190e-03, 7.2021e-03, ..., 1.2398e-04, + 1.2207e-04, 1.2112e-04], + [-5.8838e-02, -5.7617e-02, -5.6641e-02, ..., -1.5640e-04, + -1.9741e-04, -2.4986e-04], + [-7.3547e-03, -7.2327e-03, -7.4158e-03, ..., -1.4496e-04, + -1.3638e-04, -7.8678e-05], + ..., + [-3.7231e-03, -4.0283e-03, -4.3335e-03, ..., -1.0157e-04, + -2.8610e-04, -1.6403e-04], + [-5.6641e-02, -5.5420e-02, -5.4199e-02, ..., -1.6499e-04, + -9.9182e-05, -1.9646e-04], + [-7.4768e-03, -8.6060e-03, -8.1177e-03, ..., -2.5749e-04, + -1.3828e-04, -1.4877e-04]], dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0034, -0.0026, -0.0031, ..., -0.0024, -0.0036, -0.0036], + [-0.0036, -0.0008, 0.0034, ..., 0.0035, -0.0003, 0.0035], + [ 0.0013, -0.0031, -0.0013, ..., -0.0016, -0.0030, -0.0013], + ..., + [ 0.0173, -0.0244, -0.0171, ..., -0.0161, -0.0195, -0.0171], + [ 0.0112, -0.0156, -0.0111, ..., -0.0106, -0.0131, -0.0111], + [-0.0187, 0.0193, 0.0181, ..., 0.0178, 0.0187, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.1.weight': tensor([[-0.0588, -0.0579, -0.0564, ..., 0.0002, 0.0003, 0.0002], + [ 0.0713, 0.0698, 0.0674, ..., -0.0006, -0.0006, -0.0006], + [ 0.0693, 0.0679, 0.0664, ..., -0.0004, -0.0002, -0.0002], + ..., + [ 0.0334, 0.0327, 0.0317, ..., -0.0003, -0.0001, -0.0002], + [ 0.1211, 0.1177, 0.1143, ..., 0.0002, 0.0002, 0.0002], + [ 0.0728, 0.0703, 0.0693, ..., -0.0003, -0.0003, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.13.adaln_modulation_mlp.2.weight': tensor([[-7.2479e-04, 5.5542e-03, 3.9978e-03, ..., 5.3406e-03, + -3.7842e-03, 2.9907e-03], + [ 3.4485e-03, 2.6093e-03, -3.9368e-03, ..., -3.7689e-03, + -1.8311e-03, -3.5858e-03], + [-1.2589e-03, -5.1022e-05, 3.0060e-03, ..., 2.7466e-03, + -5.5237e-03, 1.3809e-03], + ..., + [ 1.8311e-02, 2.9541e-02, -7.0190e-03, ..., -1.9073e-03, + 5.0659e-03, -1.2756e-02], + [ 2.6733e-02, 2.7954e-02, -2.1362e-02, ..., -1.5198e-02, + -3.6240e-04, -2.4292e-02], + [-4.6631e-02, -3.1494e-02, 4.8828e-02, ..., 4.3701e-02, + 9.4604e-03, 5.1025e-02]], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_proj.weight': tensor([[ 0.0288, -0.0120, 0.0091, ..., 0.0061, -0.0008, 0.0176], + [ 0.0053, 0.0063, 0.0056, ..., 0.0270, 0.0025, -0.0203], + [-0.0063, 0.0181, -0.0033, ..., -0.0132, -0.0162, -0.0084], + ..., + [ 0.0048, -0.0109, -0.0053, ..., 0.0117, 0.0066, -0.0047], + [ 0.0187, -0.0137, -0.0038, ..., -0.0004, -0.0114, -0.0298], + [ 0.0039, 0.0125, 0.0248, ..., -0.0023, 0.0050, 0.0011]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm.weight': tensor([1.0703, 1.2891, 1.0938, 1.2656, 1.2031, 0.6758, 0.9531, 1.2031, 1.2656, + 0.9961, 1.1953, 1.3516, 1.1719, 1.0781, 1.3359, 1.1484, 1.2656, 1.3359, + 1.3672, 1.3828, 1.2578, 1.3438, 1.1094, 1.2109, 1.3203, 1.6641, 1.2188, + 1.2578, 1.3984, 1.1328, 1.3750, 1.2812, 0.1582, 1.3828, 1.3438, 1.2812, + 1.2734, 1.3438, 1.2891, 1.0625, 1.3125, 0.3320, 1.3125, 0.9570, 1.0781, + 0.8242, 1.2578, 1.4375, 0.8672, 0.3535, 1.3672, 1.3359, 1.2969, 1.2188, + 1.2734, 1.3594, 1.1094, 1.2578, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.4141, 1.4219, 1.2891, 1.5469, 0.9297, 0.5117, 0.4434, 0.8594, + 1.1719, 1.3281, 1.3281, 1.1797, 1.0078, 1.1797, 1.3672, 1.2891, 1.2188, + 1.1406, 0.9531, 1.3438, 1.1641, 1.2734, 1.3125, 0.9688, 1.2188, 1.3047, + 1.3359, 1.5000, 1.4688, 1.4141, 1.4141, 1.4453, 1.3281, 1.3359, 1.2031, + 1.0156, 1.2109, 0.8945, 1.3359, 1.2656, 1.1328, 1.0625, 1.2656, 1.2266, + 0.8359, 1.0781, 1.1484, 1.5000, 1.0469, 0.4941, 1.3203, 1.4844, 1.2969, + 1.3750, 1.3906, 1.3516, 1.2812, 1.2578, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.k_proj.weight': tensor([[ 0.0021, 0.0052, -0.0071, ..., -0.0054, -0.0078, 0.0105], + [ 0.0170, 0.0075, 0.0086, ..., 0.0073, -0.0106, 0.0067], + [ 0.0093, -0.0064, -0.0015, ..., 0.0075, 0.0028, 0.0081], + ..., + [-0.0403, -0.0243, 0.0036, ..., 0.0014, 0.0028, 0.0095], + [ 0.0271, 0.0142, 0.0042, ..., 0.0334, -0.0067, 0.0098], + [ 0.0076, 0.0266, 0.0220, ..., 0.0236, -0.0107, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm.weight': tensor([1.2500, 1.2344, 0.9844, 1.3047, 1.2109, 0.7734, 0.9688, 1.1484, 1.2734, + 0.9922, 1.1875, 1.3516, 1.1641, 1.0938, 1.3281, 1.1484, 1.2812, 1.3359, + 1.3672, 1.3906, 1.2578, 1.3438, 1.4766, 1.8281, 1.4688, 1.7344, 1.0703, + 1.4062, 1.4219, 1.1250, 1.3828, 1.4844, 0.1611, 1.3828, 1.3438, 1.2891, + 1.2734, 1.3516, 1.2891, 1.0703, 1.3125, 0.3320, 1.3125, 1.2266, 1.4922, + 1.6953, 1.7969, 1.6094, 0.9570, 0.3008, 1.4609, 1.3672, 1.2656, 1.2188, + 1.2734, 1.3672, 1.1172, 1.2500, 1.2734, 1.2969, 1.2266, 1.2266, 1.3516, + 1.3438, 1.5781, 1.3516, 1.3828, 1.5938, 0.8750, 0.3047, 0.4707, 0.9023, + 1.1641, 1.3281, 1.3359, 1.1641, 1.0156, 1.1719, 1.3672, 1.2891, 1.2188, + 1.1328, 0.9531, 1.3516, 1.1641, 1.2656, 1.6250, 1.6875, 1.6016, 1.5625, + 1.3828, 1.3984, 1.5078, 1.3984, 1.3906, 1.3906, 1.3359, 1.3359, 1.2031, + 1.0000, 1.2109, 0.8984, 1.3359, 1.2656, 1.1328, 1.0625, 1.2734, 1.5312, + 1.4688, 1.9219, 1.7188, 1.6641, 0.9688, 0.5000, 1.4531, 1.4219, 1.3438, + 1.3828, 1.3906, 1.3516, 1.2734, 1.2500, 1.3281, 1.2734, 1.2891, 1.3516, + 1.1562, 1.3359], dtype=torch.bfloat16), 'net.blocks.14.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.self_attn.v_proj.weight': tensor([[ 0.0134, -0.0483, 0.0104, ..., -0.0031, 0.0059, -0.0034], + [-0.0106, 0.0031, 0.0030, ..., -0.0201, -0.0066, 0.0080], + [ 0.0181, -0.0008, 0.0063, ..., 0.0067, -0.0036, -0.0121], + ..., + [-0.0064, 0.0273, 0.0175, ..., 0.0165, 0.0034, 0.0064], + [ 0.0071, 0.0229, -0.0190, ..., -0.0012, -0.0023, -0.0243], + [ 0.0199, -0.0262, -0.0215, ..., -0.0049, 0.0095, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.self_attn.output_proj.weight': tensor([[ 0.0100, -0.0253, 0.0010, ..., -0.0060, -0.0036, 0.0085], + [-0.0128, 0.0043, -0.0315, ..., -0.0254, 0.0051, 0.0025], + [ 0.0187, -0.0066, 0.0201, ..., 0.0025, -0.0228, -0.0300], + ..., + [-0.0226, 0.0082, -0.0371, ..., -0.0054, 0.0069, 0.0187], + [ 0.0248, -0.0068, -0.0097, ..., 0.0019, -0.0074, 0.0303], + [ 0.0130, -0.0201, 0.0092, ..., -0.0153, 0.0142, 0.0027]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_proj.weight': tensor([[ 0.0061, -0.0109, 0.0081, ..., 0.0043, -0.0066, 0.0099], + [-0.0208, 0.0102, -0.0157, ..., 0.0022, 0.0021, -0.0080], + [-0.0041, 0.0144, 0.0090, ..., -0.0135, -0.0116, 0.0055], + ..., + [ 0.0042, 0.0084, -0.0038, ..., -0.0118, 0.0008, -0.0239], + [ 0.0148, -0.0029, 0.0060, ..., 0.0101, 0.0027, -0.0210], + [ 0.0045, -0.0095, 0.0049, ..., -0.0023, 0.0016, 0.0188]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2676, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9609, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1367, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0547, 0.2119, 0.8555, 1.0547, 1.0469, 0.9727, + 0.3652, 0.2695, 0.3691, 0.2080, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6367, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2168, 0.5742, 0.7188, + 0.2227, 0.1465, 1.0078, 0.2793, 1.1094, 0.9961, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3047, 1.0000, 0.2988, 1.0234, 0.1367, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.8984, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2578, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0718, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_proj.weight': tensor([[-0.0013, -0.0028, -0.0056, ..., -0.0019, -0.0040, -0.0019], + [ 0.0156, 0.0134, -0.0098, ..., 0.0057, -0.0075, 0.0115], + [-0.0030, -0.0066, 0.0197, ..., 0.0014, 0.0022, -0.0129], + ..., + [-0.0046, -0.0040, 0.0009, ..., 0.0016, -0.0050, -0.0025], + [-0.0027, 0.0138, 0.0036, ..., -0.0058, -0.0013, -0.0009], + [ 0.0072, -0.0057, -0.0009, ..., -0.0043, -0.0031, -0.0037]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm.weight': tensor([0.1055, 0.3457, 0.9336, 1.0000, 0.9688, 0.3047, 0.5273, 0.3594, 0.1953, + 0.3047, 1.0078, 0.9414, 0.1572, 0.2695, 0.9258, 1.0234, 0.9531, 0.7891, + 1.0312, 0.9648, 1.0469, 1.0625, 0.4902, 0.8672, 1.0781, 0.9609, 0.1001, + 0.1357, 0.2559, 0.9961, 1.0625, 1.0469, 0.3301, 0.5781, 0.8242, 0.2275, + 1.0078, 1.0547, 0.8086, 1.0469, 0.2119, 0.8555, 1.0547, 1.0625, 0.9727, + 0.3652, 0.2676, 0.3691, 0.2070, 0.2266, 1.0234, 0.2100, 0.3066, 1.0781, + 1.0781, 0.9023, 0.9531, 1.0625, 1.0234, 0.4688, 1.0234, 0.8477, 0.6328, + 0.4492, 0.2246, 0.7383, 0.5273, 1.0234, 0.8945, 0.2158, 0.5742, 0.7188, + 0.2227, 0.1475, 1.0078, 0.2793, 1.1094, 1.0000, 0.6914, 0.3965, 0.9453, + 1.0547, 0.4961, 0.2217, 0.3027, 1.0234, 0.2988, 1.0234, 0.1357, 1.0078, + 0.3477, 0.9922, 0.2695, 0.9648, 0.2617, 0.9023, 0.2393, 1.1016, 1.1172, + 0.6367, 0.1553, 0.2930, 0.1543, 0.2129, 1.0469, 0.2598, 0.1729, 0.8047, + 1.0234, 0.8516, 0.2578, 0.9922, 1.0234, 0.6953, 0.1504, 0.2578, 0.9766, + 0.9375, 1.0547, 0.9531, 0.1641, 0.2988, 0.9180, 1.0391, 0.9648, 0.1855, + 0.0728, 0.0889], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.v_proj.weight': tensor([[-0.0251, -0.0019, 0.0045, ..., 0.0096, 0.0217, 0.0031], + [-0.0216, -0.0028, 0.0017, ..., -0.0093, -0.0250, -0.0117], + [ 0.0371, -0.0053, 0.0079, ..., -0.0006, 0.0253, 0.0187], + ..., + [ 0.0149, 0.0023, 0.0186, ..., -0.0057, -0.0004, -0.0286], + [ 0.0249, -0.0059, 0.0078, ..., 0.0092, 0.0101, 0.0071], + [-0.0044, -0.0005, -0.0027, ..., 0.0117, 0.0098, 0.0176]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.output_proj.weight': tensor([[ 0.0056, 0.0115, -0.0030, ..., 0.0090, 0.0015, -0.0155], + [ 0.0126, 0.0093, 0.0125, ..., 0.0289, 0.0182, 0.0135], + [-0.0060, 0.0017, -0.0172, ..., -0.0107, -0.0049, 0.0036], + ..., + [ 0.0037, -0.0118, 0.0082, ..., 0.0106, 0.0123, 0.0068], + [-0.0034, -0.0054, -0.0092, ..., -0.0159, -0.0221, -0.0034], + [ 0.0029, 0.0111, 0.0019, ..., 0.0168, -0.0128, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img.weight': tensor([[ 0.0188, 0.0262, -0.0079, ..., -0.0238, -0.0099, 0.0118], + [-0.0250, -0.0162, 0.0029, ..., -0.0085, -0.0020, -0.0369], + [ 0.0019, 0.0141, 0.0074, ..., -0.0452, 0.0014, 0.0119], + ..., + [ 0.0149, -0.0140, 0.0325, ..., 0.0598, -0.0221, 0.0317], + [ 0.0199, 0.0216, 0.0371, ..., 0.0096, -0.0298, 0.0254], + [-0.0189, 0.0186, -0.0352, ..., -0.0143, 0.0039, -0.0214]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.v_img.weight': tensor([[-0.0063, -0.0327, -0.0190, ..., 0.0078, 0.0013, -0.0209], + [ 0.0713, 0.0153, -0.0053, ..., -0.0035, 0.0186, 0.0002], + [-0.0227, -0.0094, 0.0188, ..., -0.0128, 0.0193, 0.0068], + ..., + [ 0.0114, 0.0464, 0.0082, ..., 0.0248, 0.0168, -0.0040], + [ 0.0381, -0.0038, 0.0063, ..., -0.0043, -0.0113, -0.0037], + [ 0.0292, 0.0044, 0.0209, ..., -0.0064, 0.0234, 0.0123]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img.weight': tensor([[ 0.0219, -0.0049, 0.0212, ..., 0.0305, 0.0136, 0.0105], + [-0.0077, -0.0366, -0.0160, ..., 0.0317, -0.0215, -0.0243], + [-0.0278, -0.0282, -0.0199, ..., 0.0110, -0.0223, -0.0049], + ..., + [ 0.0084, -0.0275, 0.0225, ..., 0.0055, -0.0189, -0.0105], + [-0.0125, 0.0047, -0.0058, ..., 0.0038, -0.0036, 0.0211], + [-0.0061, 0.0038, -0.0060, ..., -0.0110, 0.0264, 0.0046]], + dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.14.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.14.mlp.layer1.weight': tensor([[-2.6855e-03, -1.0620e-02, 2.0752e-02, ..., -1.1047e-02, + -2.6123e-02, -3.5553e-03], + [-2.3346e-03, -6.4697e-03, 8.4229e-03, ..., -9.5825e-03, + -2.2583e-02, 3.2349e-03], + [ 3.5400e-03, 9.8267e-03, 3.6926e-03, ..., -1.0742e-02, + -2.5269e-02, 2.6584e-05], + ..., + [ 1.7944e-02, 1.7822e-02, 1.7334e-02, ..., -2.4902e-02, + 2.8381e-03, -1.0252e-05], + [ 2.4902e-02, -4.0588e-03, -2.2095e-02, ..., -8.9645e-04, + -7.3242e-03, 1.5198e-02], + [-1.4954e-02, -3.9368e-03, 3.9864e-04, ..., -2.4292e-02, + -1.7090e-03, 8.3618e-03]], dtype=torch.bfloat16), 'net.blocks.14.mlp.layer2.weight': tensor([[-0.0150, -0.0172, -0.0007, ..., -0.0081, 0.0109, -0.0038], + [-0.0187, -0.0009, 0.0028, ..., -0.0077, 0.0332, -0.0028], + [-0.0168, 0.0155, 0.0031, ..., -0.0022, -0.0195, 0.0081], + ..., + [-0.0151, -0.0082, 0.0059, ..., 0.0145, 0.0048, -0.0105], + [-0.0085, -0.0281, -0.0118, ..., 0.0022, 0.0029, -0.0037], + [ 0.0234, -0.0053, 0.0273, ..., -0.0074, -0.0214, -0.0018]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.1.weight': tensor([[ 8.3496e-02, 8.1055e-02, 7.9102e-02, ..., 2.4796e-04, + 2.2125e-04, 1.0920e-04], + [-2.7618e-03, -2.8687e-03, -2.8381e-03, ..., 9.0122e-05, + 7.2479e-05, 1.0681e-04], + [-7.3242e-02, -7.2266e-02, -7.0312e-02, ..., 3.5858e-04, + 2.3460e-04, 2.4986e-04], + ..., + [-7.2754e-02, -7.1289e-02, -7.0312e-02, ..., 2.0599e-04, + 2.0409e-04, 2.4605e-04], + [-7.0312e-02, -6.8359e-02, -6.6895e-02, ..., 1.3161e-04, + 2.8038e-04, 1.9264e-04], + [-1.0559e-02, -1.1047e-02, -1.1353e-02, ..., 1.4210e-04, + 8.3447e-05, 1.3256e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_self_attn.2.weight': tensor([[ 0.0118, -0.0003, 0.0146, ..., 0.0101, 0.0090, -0.0002], + [ 0.0042, -0.0032, 0.0024, ..., 0.0004, 0.0002, -0.0029], + [-0.0001, -0.0043, -0.0029, ..., -0.0025, -0.0033, -0.0043], + ..., + [ 0.0085, 0.0203, 0.0064, ..., 0.0104, 0.0119, 0.0205], + [ 0.0104, 0.0220, 0.0211, ..., 0.0216, 0.0216, 0.0226], + [ 0.0017, -0.0310, -0.0210, ..., -0.0251, -0.0264, -0.0312]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.1.weight': tensor([[ 0.0388, 0.0378, 0.0374, ..., 0.0002, 0.0002, 0.0001], + [ 0.0123, 0.0118, 0.0124, ..., 0.0003, 0.0001, 0.0002], + [-0.0046, -0.0045, -0.0052, ..., -0.0001, -0.0002, -0.0002], + ..., + [-0.0386, -0.0378, -0.0374, ..., -0.0002, -0.0002, -0.0002], + [ 0.0047, 0.0045, 0.0052, ..., 0.0002, 0.0002, 0.0001], + [-0.0120, -0.0117, -0.0124, ..., -0.0002, -0.0002, -0.0001]], + dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_cross_attn.2.weight': tensor([[-8.6670e-03, -7.9956e-03, 7.6599e-03, ..., 8.6060e-03, + -7.7209e-03, 7.7820e-03], + [-4.2725e-04, 1.3542e-04, -2.5940e-04, ..., 4.7302e-04, + 2.3365e-04, -1.4019e-04], + [-2.3842e-04, 2.5177e-04, -2.3270e-04, ..., 2.1267e-04, + 4.2343e-04, -9.5367e-05], + ..., + [ 2.1973e-02, 2.0752e-02, -2.0020e-02, ..., -2.1851e-02, + 1.9775e-02, -2.0752e-02], + [-2.9663e-02, -2.9053e-02, 2.9053e-02, ..., 2.9663e-02, + -2.8442e-02, 2.8809e-02], + [-1.9165e-02, -1.9165e-02, 1.8433e-02, ..., 1.8799e-02, + -1.8311e-02, 1.8921e-02]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.1.weight': tensor([[-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., 1.0777e-04, + 1.8024e-04, 2.2888e-04], + [-1.1182e-01, -1.0840e-01, -1.0547e-01, ..., -1.8311e-04, + -1.2302e-04, -1.2302e-04], + [ 1.2891e-01, 1.2598e-01, 1.2354e-01, ..., -1.8978e-04, + -1.7262e-04, -1.6403e-04], + ..., + [-6.9336e-02, -6.7383e-02, -6.5918e-02, ..., -1.4019e-04, + -1.7452e-04, -1.3447e-04], + [-7.0312e-02, -6.8848e-02, -6.6895e-02, ..., 2.7084e-04, + 2.4128e-04, 1.9836e-04], + [-4.6387e-02, -4.5166e-02, -4.4189e-02, ..., 2.2888e-04, + 3.0899e-04, 1.4400e-04]], dtype=torch.bfloat16), 'net.blocks.14.adaln_modulation_mlp.2.weight': tensor([[ 5.0659e-03, 1.0071e-02, -3.6163e-03, ..., 1.0010e-02, + -3.6469e-03, -2.8992e-04], + [-1.0300e-04, 8.3447e-05, 1.7738e-04, ..., -5.0354e-04, + 1.3351e-03, 2.9297e-03], + [ 7.3242e-03, 5.3101e-03, -8.3618e-03, ..., 6.6833e-03, + -1.0757e-03, -3.2196e-03], + ..., + [-1.3885e-03, 3.4332e-03, 1.3580e-03, ..., 5.7602e-04, + 1.6113e-02, 2.3682e-02], + [ 9.2163e-03, 2.7161e-03, -8.4839e-03, ..., -8.6670e-03, + 3.6865e-02, 3.5400e-02], + [-1.8311e-02, -1.5625e-02, 1.6968e-02, ..., -2.2602e-04, + -4.4678e-02, -4.5410e-02]], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_proj.weight': tensor([[ 0.0118, -0.0008, 0.0060, ..., -0.0109, 0.0071, -0.0189], + [-0.0308, 0.0032, 0.0084, ..., -0.0134, 0.0221, -0.0154], + [ 0.0115, 0.0028, -0.0004, ..., 0.0049, -0.0129, 0.0114], + ..., + [-0.0272, 0.0023, -0.0147, ..., 0.0014, 0.0162, 0.0042], + [ 0.0063, -0.0255, 0.0051, ..., 0.0153, 0.0330, -0.0164], + [ 0.0034, -0.0223, -0.0048, ..., -0.0041, 0.0108, -0.0041]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm.weight': tensor([1.3359, 1.2656, 0.9297, 1.6094, 0.8359, 1.0312, 1.4609, 1.3828, 1.4609, + 1.4219, 1.4844, 1.5312, 1.4375, 1.4453, 1.4453, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9883, 1.5156, 1.2812, 1.2344, 0.3164, + 0.2695, 0.3281, 1.4453, 1.4219, 1.4766, 1.5078, 1.2578, 1.3750, 1.4844, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9453, 1.2266, + 1.3516, 1.2656, 1.5547, 0.3359, 0.8320, 0.8438, 0.8164, 1.4609, 1.5234, + 1.2969, 1.4609, 1.4609, 1.3984, 1.4297, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.2266, 1.1719, 1.0859, 1.2891, 0.7578, 0.7891, 0.8789, 1.3984, + 1.4219, 1.4062, 1.4609, 1.4141, 1.4766, 1.4141, 1.5547, 1.4375, 0.9727, + 1.4375, 1.4531, 0.8711, 1.4297, 1.1016, 1.0938, 1.1641, 1.3828, 1.2344, + 0.7305, 0.4805, 0.6875, 0.8516, 1.2109, 1.2812, 1.4219, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4609, 1.5312, 1.4844, 0.9883, + 1.3906, 0.9961, 1.3047, 1.1250, 0.7383, 0.7969, 0.7383, 1.5859, 1.4609, + 1.3672, 1.4453, 1.3750, 1.5000, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.k_proj.weight': tensor([[-0.0085, -0.0024, -0.0054, ..., -0.0304, 0.0076, -0.0186], + [-0.0154, 0.0042, 0.0011, ..., 0.0031, -0.0162, -0.0045], + [ 0.0272, 0.0009, -0.0096, ..., -0.0090, -0.0154, 0.0084], + ..., + [ 0.0124, -0.0039, 0.0008, ..., 0.0048, 0.0150, -0.0091], + [ 0.0093, -0.0033, 0.0095, ..., -0.0096, -0.0054, -0.0006], + [-0.0025, -0.0171, -0.0056, ..., -0.0239, 0.0410, 0.0042]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm.weight': tensor([1.2188, 0.9609, 0.4355, 1.4453, 0.8750, 0.9883, 1.3828, 1.3750, 1.4609, + 1.4219, 1.4609, 1.5078, 1.4375, 1.4453, 1.4609, 1.5156, 1.4531, 1.4609, + 1.2656, 1.4922, 1.4297, 1.4922, 0.9258, 1.7109, 1.6172, 1.6172, 0.5547, + 1.0078, 0.9297, 1.2188, 1.4844, 1.5078, 1.4844, 1.2500, 1.3750, 1.4766, + 1.4062, 1.4922, 1.4766, 1.4609, 1.4922, 1.2266, 1.2578, 0.9297, 1.3828, + 1.3672, 1.3828, 1.7188, 0.4551, 1.1328, 0.9844, 1.0312, 1.4844, 1.5234, + 1.2812, 1.4609, 1.4609, 1.3906, 1.4375, 1.4922, 1.4922, 1.4688, 1.4609, + 1.4609, 1.1016, 1.3906, 1.1875, 1.2656, 0.6328, 0.8320, 1.2812, 1.3672, + 1.3984, 1.3906, 1.4688, 1.4531, 1.4844, 1.4141, 1.5547, 1.4375, 0.9688, + 1.4375, 1.4453, 0.8711, 1.4297, 1.1016, 1.0781, 1.5547, 1.3281, 1.3984, + 0.7266, 0.5156, 0.8516, 1.0391, 1.2578, 1.2500, 1.4531, 1.4375, 1.3125, + 1.4531, 1.4531, 1.3828, 1.4766, 1.5234, 1.4531, 1.5312, 1.4844, 1.0078, + 1.2188, 1.0625, 1.4609, 1.2734, 0.7617, 1.0078, 0.6641, 1.4219, 1.4375, + 1.3750, 1.4531, 1.3750, 1.5078, 1.5078, 1.5156, 1.4688, 1.4688, 1.4766, + 1.4219, 1.2109], dtype=torch.bfloat16), 'net.blocks.15.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.self_attn.v_proj.weight': tensor([[-0.0097, 0.0200, 0.0004, ..., 0.0141, 0.0097, 0.0065], + [ 0.0106, 0.0187, -0.0186, ..., -0.0015, 0.0146, 0.0001], + [-0.0048, 0.0094, 0.0111, ..., 0.0077, 0.0100, 0.0205], + ..., + [ 0.0009, -0.0095, -0.0096, ..., 0.0090, 0.0140, 0.0104], + [ 0.0097, 0.0051, 0.0100, ..., 0.0175, -0.0156, 0.0126], + [ 0.0047, 0.0037, 0.0178, ..., 0.0036, -0.0240, -0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.self_attn.output_proj.weight': tensor([[ 0.0047, 0.0284, -0.0189, ..., 0.0092, -0.0135, 0.0059], + [-0.0275, -0.0029, 0.0025, ..., -0.0143, -0.0128, -0.0055], + [ 0.0091, 0.0027, 0.0112, ..., 0.0030, 0.0022, -0.0072], + ..., + [-0.0154, -0.0038, 0.0018, ..., 0.0069, -0.0054, 0.0027], + [-0.0049, -0.0066, 0.0236, ..., 0.0097, -0.0317, -0.0104], + [-0.0042, 0.0050, -0.0160, ..., -0.0186, -0.0042, 0.0187]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_proj.weight': tensor([[-0.0110, -0.0117, 0.0055, ..., -0.0042, -0.0148, -0.0164], + [ 0.0178, 0.0190, -0.0190, ..., 0.0015, -0.0137, 0.0145], + [ 0.0050, -0.0232, 0.0055, ..., 0.0080, 0.0081, -0.0110], + ..., + [-0.0115, -0.0094, 0.0104, ..., 0.0132, 0.0232, 0.0106], + [ 0.0038, 0.0325, 0.0128, ..., 0.0043, -0.0312, 0.0189], + [-0.0155, 0.0183, -0.0096, ..., 0.0023, -0.0184, -0.0234]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm.weight': tensor([1.2109, 1.1172, 0.8711, 0.3828, 0.0947, 0.9336, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9883, 0.3496, 0.9336, 0.1592, 1.1250, 0.1426, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1250, 0.0596, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5820, 1.0703, 1.2188, 0.0815, 1.1797, 0.2178, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1602, 0.2236, 0.8711, 0.6719, 1.1953, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1562, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0693, 1.0312, 0.5703, 0.2012, 0.0559, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9141, + 1.1172, 1.0156, 1.0547, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_proj.weight': tensor([[ 0.0009, 0.0026, 0.0172, ..., -0.0032, 0.0211, 0.0036], + [-0.0007, 0.0011, 0.0007, ..., 0.0070, 0.0221, 0.0082], + [-0.0068, 0.0156, 0.0042, ..., 0.0001, -0.0152, -0.0131], + ..., + [-0.0129, 0.0013, 0.0167, ..., -0.0171, 0.0172, -0.0295], + [ 0.0264, -0.0049, -0.0045, ..., -0.0184, -0.0173, 0.0149], + [ 0.0165, 0.0030, -0.0009, ..., -0.0223, 0.0028, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm.weight': tensor([1.2109, 1.1172, 0.8672, 0.3828, 0.0947, 0.9297, 0.5703, 0.8828, 1.0625, + 0.9922, 1.1875, 0.3301, 0.1221, 1.0859, 0.7617, 0.2852, 1.1016, 1.0391, + 0.7891, 0.9766, 0.9922, 0.3496, 0.9375, 0.1592, 1.1250, 0.1416, 1.1641, + 1.1016, 1.1406, 1.1797, 1.1484, 0.2324, 0.9414, 1.0938, 1.1484, 0.0972, + 1.1719, 1.1641, 1.0625, 1.1641, 1.1328, 0.0591, 0.7812, 1.1406, 1.1172, + 0.9141, 0.5859, 1.0703, 1.2109, 0.0815, 1.1797, 0.2158, 1.1406, 1.0781, + 0.1963, 1.0938, 1.0938, 1.1641, 1.1562, 1.2031, 0.0996, 0.2910, 1.1641, + 0.1621, 1.0781, 1.1484, 1.0156, 0.1611, 0.2236, 0.8711, 0.6680, 1.1875, + 1.1016, 1.0078, 1.1641, 1.0156, 0.2910, 0.3203, 1.2344, 0.2949, 0.5938, + 1.0625, 0.8125, 0.2314, 0.5117, 1.1484, 0.8984, 0.3320, 1.2344, 0.8828, + 0.3633, 1.1484, 0.4082, 0.2676, 1.1016, 0.9609, 1.2266, 0.6328, 1.1875, + 0.2002, 0.2041, 0.0698, 1.0234, 0.5703, 0.2012, 0.0554, 0.7539, 0.4395, + 0.7148, 0.9609, 0.9375, 1.1328, 1.0547, 1.0234, 1.0781, 0.2871, 0.9180, + 1.1172, 1.0156, 1.0469, 1.1328, 1.1406, 1.0859, 0.8594, 1.1250, 1.1484, + 1.1406, 1.1016], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.v_proj.weight': tensor([[ 0.0003, 0.0005, -0.0017, ..., -0.0020, 0.0061, 0.0228], + [-0.0181, -0.0015, -0.0022, ..., -0.0032, -0.0586, 0.0138], + [ 0.0033, 0.0050, -0.0034, ..., 0.0144, 0.0026, 0.0130], + ..., + [ 0.0219, -0.0059, 0.0023, ..., -0.0112, -0.0306, 0.0238], + [ 0.0238, -0.0102, -0.0149, ..., 0.0048, 0.0076, -0.0054], + [ 0.0115, 0.0095, -0.0164, ..., -0.0049, -0.0135, 0.0095]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.output_proj.weight': tensor([[-0.0190, 0.0127, 0.0190, ..., 0.0162, 0.0179, -0.0045], + [ 0.0079, 0.0068, 0.0045, ..., -0.0120, 0.0009, 0.0177], + [ 0.0059, 0.0038, 0.0119, ..., 0.0037, 0.0036, -0.0020], + ..., + [-0.0080, -0.0095, -0.0044, ..., 0.0240, -0.0135, 0.0145], + [ 0.0259, -0.0134, 0.0149, ..., 0.0179, -0.0189, -0.0105], + [-0.0054, -0.0071, 0.0243, ..., 0.0292, -0.0175, 0.0183]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img.weight': tensor([[ 4.6387e-03, -7.4768e-03, -1.0864e-02, ..., 1.8555e-02, + -2.1606e-02, 5.0964e-03], + [-3.8300e-03, 1.2207e-02, -2.3071e-02, ..., 1.2329e-02, + -1.1230e-02, -1.8311e-02], + [ 1.0681e-02, 1.0315e-02, 3.9041e-06, ..., 4.2480e-02, + 3.0884e-02, -2.7344e-02], + ..., + [ 3.3203e-02, 1.2390e-02, 1.4587e-02, ..., 5.3955e-02, + 4.1504e-02, 9.7046e-03], + [ 3.8605e-03, 8.5449e-03, -1.4038e-02, ..., 4.4250e-03, + -1.8921e-02, -2.3682e-02], + [-4.3030e-03, 3.5400e-02, 1.7700e-02, ..., 3.6133e-02, + -3.1128e-02, 6.9885e-03]], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.v_img.weight': tensor([[ 0.0500, -0.0061, 0.0359, ..., -0.0090, -0.0128, -0.0022], + [ 0.0190, -0.0052, 0.0298, ..., 0.0125, -0.0043, -0.0200], + [ 0.0116, -0.0109, 0.0177, ..., -0.0248, 0.0229, 0.0014], + ..., + [-0.0082, 0.0022, -0.0396, ..., -0.0198, -0.0220, 0.0591], + [-0.0317, -0.0047, -0.0291, ..., 0.0206, 0.0304, -0.0090], + [-0.0072, 0.0245, 0.0047, ..., -0.0413, 0.0216, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img.weight': tensor([[-0.0062, -0.0157, -0.0065, ..., -0.0067, 0.0228, -0.0339], + [-0.0713, -0.0156, -0.0168, ..., -0.0217, -0.0183, -0.0069], + [-0.0201, -0.0157, -0.0167, ..., -0.0175, 0.0238, -0.0038], + ..., + [ 0.0046, -0.0049, 0.0258, ..., -0.0310, 0.0090, -0.0361], + [ 0.0061, 0.0308, 0.0542, ..., -0.0542, -0.0155, 0.0259], + [-0.0022, 0.0157, -0.0074, ..., -0.0018, -0.0249, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.15.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.15.mlp.layer1.weight': tensor([[ 2.3804e-02, -1.6861e-03, -1.7090e-02, ..., 3.4790e-03, + -7.7515e-03, 1.6479e-02], + [-1.4954e-02, -4.1389e-04, 8.3618e-03, ..., 1.6479e-02, + 4.1504e-03, 2.3499e-03], + [ 1.3855e-02, 1.7643e-05, 3.8605e-03, ..., -1.1169e-02, + 7.2632e-03, 3.0518e-03], + ..., + [ 1.6479e-02, 5.5542e-03, -6.3477e-03, ..., 3.4637e-03, + -1.3855e-02, 3.3447e-02], + [ 2.9144e-03, 1.9287e-02, -2.8076e-02, ..., 1.4160e-02, + 1.3550e-02, -1.3489e-02], + [ 1.7578e-02, -2.5787e-03, 5.6458e-03, ..., -1.4572e-03, + -1.3245e-02, 2.3193e-02]], dtype=torch.bfloat16), 'net.blocks.15.mlp.layer2.weight': tensor([[-0.0118, -0.0181, 0.0003, ..., 0.0054, 0.0038, 0.0132], + [ 0.0052, -0.0186, -0.0053, ..., -0.0087, 0.0123, -0.0053], + [ 0.0010, 0.0145, -0.0089, ..., 0.0067, 0.0038, -0.0309], + ..., + [-0.0009, -0.0047, -0.0053, ..., 0.0315, 0.0049, 0.0074], + [-0.0161, -0.0102, 0.0161, ..., -0.0128, 0.0043, 0.0046], + [ 0.0154, -0.0063, -0.0003, ..., -0.0258, -0.0030, 0.0408]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.1.weight': tensor([[-1.0498e-01, -1.0303e-01, -1.0059e-01, ..., -1.8978e-04, + -1.3161e-04, -1.6022e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.4796e-04, + -3.0708e-04, -2.5558e-04], + [ 6.7871e-02, 6.5918e-02, 6.4941e-02, ..., -2.6703e-04, + -2.8610e-04, -3.8719e-04], + ..., + [-1.3184e-01, -1.2793e-01, -1.2598e-01, ..., -1.4019e-04, + -7.3910e-05, -7.8201e-05], + [-6.1523e-02, -6.0059e-02, -5.8350e-02, ..., 4.0627e-04, + 3.1090e-04, 3.9101e-04], + [-6.8359e-02, -6.6895e-02, -6.5430e-02, ..., 3.2234e-04, + 3.1090e-04, 3.0136e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_self_attn.2.weight': tensor([[ 0.0149, -0.0055, -0.0060, ..., 0.0193, 0.0064, 0.0057], + [ 0.0070, -0.0015, -0.0021, ..., 0.0146, 0.0029, 0.0015], + [ 0.0036, 0.0055, 0.0065, ..., -0.0002, -0.0070, -0.0064], + ..., + [-0.0198, -0.0101, -0.0075, ..., -0.0195, 0.0032, 0.0100], + [ 0.0039, -0.0139, -0.0115, ..., 0.0053, 0.0025, 0.0128], + [-0.0109, 0.0405, 0.0420, ..., -0.0219, -0.0400, -0.0417]], + dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.1.weight': tensor([[ 8.5449e-04, -1.6174e-03, -2.3346e-03, ..., 1.3351e-04, + 4.3154e-05, 1.2875e-04], + [-2.7588e-02, -2.7344e-02, -2.6978e-02, ..., 9.4414e-05, + 6.7711e-05, 1.8311e-04], + [-1.8692e-03, -9.9182e-04, -9.1171e-04, ..., -8.9645e-05, + -8.4877e-05, -1.1015e-04], + ..., + [ 8.5068e-04, 8.5831e-04, -1.6174e-03, ..., 1.5545e-04, + 4.2915e-05, 1.4210e-04], + [ 5.6458e-03, 4.5776e-03, 4.3640e-03, ..., 1.5450e-04, + 6.1989e-05, 2.2888e-04], + [ 4.6082e-03, 3.5095e-03, 3.4485e-03, ..., 1.4210e-04, + 1.3638e-04, 1.5831e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_cross_attn.2.weight': tensor([[-2.6855e-03, -3.8605e-03, 1.9836e-03, ..., -2.7161e-03, + -5.8746e-04, -6.3324e-04], + [-1.6632e-03, -2.9602e-03, 2.1667e-03, ..., -1.6479e-03, + -2.5024e-03, -2.3346e-03], + [ 1.9684e-03, -9.8705e-05, -2.4261e-03, ..., 1.7929e-03, + 2.9144e-03, 2.8381e-03], + ..., + [-2.1851e-02, -2.2217e-02, 2.0020e-02, ..., -2.1729e-02, + -1.8188e-02, -1.8677e-02], + [-2.4658e-02, -2.4536e-02, 2.3315e-02, ..., -2.4536e-02, + -2.2461e-02, -2.2583e-02], + [ 3.1982e-02, 3.2471e-02, -3.0273e-02, ..., 3.1738e-02, + 2.9785e-02, 2.9785e-02]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.1.weight': tensor([[ 6.3477e-03, 5.5237e-03, 4.6387e-03, ..., 9.8705e-05, + 6.0081e-05, 8.1539e-05], + [-6.3171e-03, -5.4321e-03, -4.7913e-03, ..., -9.7275e-05, + -1.4877e-04, -1.6308e-04], + [ 7.0801e-02, 6.7871e-02, 6.6406e-02, ..., -1.6880e-04, + -2.3651e-04, -1.2302e-04], + ..., + [ 5.7129e-02, 5.5664e-02, 5.4688e-02, ..., -1.6785e-04, + -1.4305e-04, -1.4305e-04], + [-6.4941e-02, -6.3477e-02, -6.1523e-02, ..., -1.8024e-04, + -1.9264e-04, -1.8597e-04], + [ 7.3730e-02, 7.1289e-02, 7.0801e-02, ..., -9.9659e-05, + -1.2112e-04, -1.6403e-04]], dtype=torch.bfloat16), 'net.blocks.15.adaln_modulation_mlp.2.weight': tensor([[ 5.4932e-04, -6.1798e-04, -1.2338e-05, ..., -6.0272e-04, + 1.3123e-02, -8.6975e-04], + [ 1.2817e-03, -1.3351e-03, -5.4169e-04, ..., -1.2512e-03, + -5.0049e-03, -1.0986e-03], + [ 3.4714e-04, -5.3024e-04, -1.7738e-04, ..., 1.2970e-04, + 9.2163e-03, -3.6240e-04], + ..., + [ 3.2227e-02, -3.2227e-02, -1.3916e-02, ..., -2.6123e-02, + 1.3062e-02, -2.2095e-02], + [ 3.7109e-02, -3.7109e-02, -3.5400e-02, ..., -3.7598e-02, + 1.0925e-02, -3.9062e-02], + [-4.4189e-02, 4.4189e-02, 5.2979e-02, ..., 4.9805e-02, + 1.0132e-02, 5.2002e-02]], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_proj.weight': tensor([[-0.0104, 0.0036, -0.0031, ..., -0.0044, 0.0031, 0.0084], + [ 0.0130, 0.0130, 0.0001, ..., -0.0047, 0.0037, -0.0028], + [-0.0078, -0.0015, 0.0052, ..., -0.0140, 0.0240, -0.0042], + ..., + [ 0.0036, -0.0176, -0.0034, ..., -0.0286, 0.0018, 0.0179], + [ 0.0044, 0.0035, 0.0047, ..., 0.0116, 0.0060, 0.0159], + [-0.0288, 0.0106, -0.0082, ..., 0.0035, -0.0033, 0.0094]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm.weight': tensor([0.7109, 1.7031, 1.4766, 0.7773, 1.4531, 0.5547, 0.9258, 0.9375, 1.3828, + 1.2188, 1.2656, 1.2656, 1.1953, 1.2500, 1.1562, 1.2188, 1.2500, 1.0547, + 1.1641, 1.2500, 1.3438, 1.3438, 1.5547, 1.0625, 1.2422, 1.1172, 1.5625, + 1.2109, 1.1953, 1.2344, 1.2500, 1.2422, 1.2969, 1.1328, 1.1719, 1.1094, + 1.2500, 0.2695, 0.7383, 1.2500, 0.7969, 1.2422, 1.1328, 1.2266, 1.5781, + 1.4609, 1.8594, 1.3203, 1.7188, 0.9570, 1.1953, 1.1875, 1.3281, 1.2734, + 1.2812, 1.2812, 1.2266, 1.1797, 1.2031, 1.2109, 1.2031, 1.2734, 1.0547, + 1.1875, 1.7188, 0.0811, 0.7539, 1.4297, 0.6992, 0.4277, 0.7461, 1.3438, + 1.1719, 1.2109, 1.0781, 1.1797, 1.1484, 1.0859, 1.1797, 1.1875, 1.1094, + 1.2422, 1.1562, 1.1875, 1.2734, 1.1719, 0.9609, 1.6719, 1.6562, 1.5000, + 1.1562, 1.1719, 1.1484, 1.3906, 1.3594, 1.2891, 1.0938, 1.2969, 1.1953, + 1.0938, 1.2188, 1.2578, 0.0820, 1.2188, 1.0703, 1.2969, 1.0547, 1.0156, + 0.6875, 1.0703, 0.9102, 1.0781, 1.2109, 0.9844, 0.9219, 1.3750, 1.1953, + 1.0781, 1.1875, 1.2812, 1.1016, 1.3203, 1.1406, 1.2266, 1.2578, 1.2578, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.k_proj.weight': tensor([[-0.0090, -0.0033, -0.0048, ..., -0.0081, 0.0081, 0.0056], + [-0.0099, 0.0170, -0.0122, ..., -0.0366, -0.0117, 0.0079], + [-0.0029, -0.0166, 0.0143, ..., -0.0046, 0.0155, -0.0006], + ..., + [-0.0120, -0.0044, 0.0134, ..., -0.0036, -0.0033, 0.0087], + [-0.0154, -0.0005, 0.0090, ..., 0.0033, 0.0101, 0.0049], + [ 0.0061, -0.0051, 0.0193, ..., 0.0198, 0.0008, -0.0228]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm.weight': tensor([0.9414, 1.3828, 1.4375, 0.5859, 1.3828, 0.5234, 0.9102, 0.9141, 1.3438, + 1.2109, 1.2656, 1.2734, 1.2031, 1.2500, 1.1562, 1.2188, 1.2578, 1.0547, + 1.1641, 1.2422, 1.3438, 1.3438, 1.3672, 1.1797, 1.4219, 1.2422, 1.4297, + 1.1797, 1.2578, 1.3438, 1.2578, 1.2344, 1.2969, 1.1328, 1.1797, 1.1016, + 1.2500, 0.2676, 0.7383, 1.2422, 0.7969, 1.2422, 1.1328, 1.5625, 1.5156, + 1.2109, 1.6406, 1.1797, 1.7266, 1.0469, 1.1562, 1.1641, 1.3125, 1.2578, + 1.2734, 1.2734, 1.2188, 1.1953, 1.2031, 1.2109, 1.2031, 1.2656, 1.0547, + 1.1875, 1.7109, 1.0391, 0.9492, 1.4688, 0.8906, 0.4473, 0.7500, 1.3516, + 1.2500, 1.2188, 1.0781, 1.1797, 1.1406, 1.0781, 1.1797, 1.1953, 1.1094, + 1.2500, 1.1641, 1.1875, 1.2734, 1.1797, 1.2969, 1.2812, 1.7500, 1.4375, + 1.3516, 1.2969, 1.2422, 1.3203, 1.3359, 1.2891, 1.0938, 1.2969, 1.1875, + 1.0938, 1.2188, 1.2656, 0.0815, 1.2188, 1.0703, 1.2969, 1.0547, 1.2734, + 1.2969, 1.3672, 1.3672, 1.2031, 1.2578, 0.8984, 1.1094, 1.3672, 1.1953, + 1.0781, 1.1875, 1.2812, 1.0938, 1.3203, 1.1562, 1.2344, 1.2578, 1.2656, + 1.1406, 1.2500], dtype=torch.bfloat16), 'net.blocks.16.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.self_attn.v_proj.weight': tensor([[-0.0112, -0.0059, -0.0112, ..., -0.0134, -0.0119, -0.0177], + [ 0.0059, 0.0112, -0.0121, ..., -0.0001, 0.0212, 0.0055], + [-0.0074, -0.0011, -0.0003, ..., -0.0056, 0.0046, 0.0002], + ..., + [ 0.0097, 0.0059, -0.0334, ..., 0.0142, 0.0142, -0.0002], + [ 0.0128, -0.0037, -0.0197, ..., 0.0192, -0.0067, 0.0135], + [ 0.0093, 0.0088, -0.0121, ..., -0.0145, -0.0075, 0.0325]], + dtype=torch.bfloat16), 'net.blocks.16.self_attn.output_proj.weight': tensor([[-1.6479e-02, -1.0010e-02, -6.3171e-03, ..., -1.1215e-03, + 5.0354e-03, -9.7656e-03], + [ 1.9775e-02, 4.2725e-03, 4.7493e-04, ..., 4.6997e-03, + -6.7749e-03, -1.6357e-02], + [-4.7302e-03, -7.9956e-03, -1.0757e-03, ..., -2.1973e-02, + -2.2949e-02, -1.2573e-02], + ..., + [ 1.1475e-02, -1.5747e-02, -1.3306e-02, ..., -1.3000e-02, + 1.2817e-02, -2.9541e-02], + [-7.3433e-05, 4.4556e-03, 7.3547e-03, ..., 3.0029e-02, + -4.1504e-03, 1.0757e-03], + [ 1.9287e-02, 3.5706e-03, 8.7280e-03, ..., -4.6997e-03, + 2.7313e-03, -2.5391e-02]], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_proj.weight': tensor([[ 0.0004, 0.0063, 0.0021, ..., 0.0002, -0.0084, 0.0104], + [-0.0141, -0.0057, 0.0117, ..., -0.0026, 0.0099, 0.0166], + [ 0.0231, -0.0028, 0.0152, ..., -0.0041, 0.0038, 0.0076], + ..., + [ 0.0049, 0.0197, -0.0118, ..., -0.0247, -0.0084, -0.0103], + [ 0.0052, 0.0139, 0.0042, ..., 0.0031, -0.0140, -0.0077], + [-0.0153, 0.0062, -0.0118, ..., 0.0054, -0.0211, 0.0189]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0757, 0.9609, 0.3379, 0.7578, 0.8789, 0.4590, 0.2480, 0.1040, 0.0300, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6836, 0.0898, 0.1875, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1216, 1.0312, 0.1465, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0381, 0.9102, 0.8281, 1.0156, 0.7617, 0.0669, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2129, 0.0933, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0479, 0.0767, 0.4941, + 0.1060, 0.6484, 0.2539, 0.1396, 1.0156, 1.0938, 1.0078, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0228, 0.1055, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0267, + 0.1475, 0.8203, 0.0339, 0.2451, 0.0649, 0.1196, 0.1309, 0.5859, 0.0425, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0312, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9922, 0.9844, 1.0312, 0.5703, 0.0732, 0.0771, 0.3457, 1.0703, + 0.4160, 0.1025, 0.9062, 0.9961, 0.7383, 0.0767, 0.3223, 0.1270, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_proj.weight': tensor([[-0.0064, -0.0029, -0.0056, ..., 0.0059, 0.0092, 0.0028], + [ 0.0146, 0.0028, -0.0092, ..., -0.0022, -0.0062, 0.0027], + [ 0.0083, 0.0128, 0.0112, ..., 0.0151, 0.0089, 0.0160], + ..., + [ 0.0063, 0.0076, -0.0041, ..., -0.0071, 0.0067, -0.0012], + [-0.0063, -0.0011, -0.0052, ..., 0.0078, 0.0082, 0.0069], + [ 0.0022, -0.0067, -0.0099, ..., 0.0105, 0.0162, -0.0164]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm.weight': tensor([0.2754, 0.2520, 1.0469, 0.5898, 0.1895, 1.0078, 0.3496, 0.0845, 0.8750, + 0.0752, 0.9609, 0.3379, 0.7578, 0.8828, 0.4590, 0.2480, 0.1040, 0.0304, + 0.0342, 0.0981, 0.5625, 0.9922, 0.6875, 0.0898, 0.1895, 0.2148, 0.7266, + 0.0645, 0.7656, 0.1221, 1.0312, 0.1484, 0.3125, 0.1611, 0.4531, 0.0947, + 0.0378, 0.9062, 0.8281, 1.0156, 0.7617, 0.0664, 0.3125, 0.0684, 0.6094, + 0.9375, 0.7891, 0.8828, 0.5586, 0.1934, 0.2139, 0.0928, 1.0156, 0.9961, + 0.9648, 0.7812, 0.2305, 0.9805, 0.9883, 0.9023, 0.0481, 0.0776, 0.4941, + 0.1069, 0.6445, 0.2539, 0.1396, 1.0156, 1.0938, 0.9961, 0.8594, 0.9609, + 0.8984, 0.9727, 0.2471, 0.0615, 0.7812, 0.1191, 1.0234, 0.9688, 0.9414, + 1.0703, 0.0236, 0.1060, 0.7656, 0.5352, 1.0234, 0.9531, 0.0530, 0.0262, + 0.1475, 0.8203, 0.0342, 0.2432, 0.0649, 0.1191, 0.1309, 0.5859, 0.0427, + 0.7891, 0.6484, 0.0457, 0.6406, 1.0234, 0.5664, 0.4121, 1.0312, 1.0078, + 0.8203, 0.9961, 0.9844, 1.0391, 0.5703, 0.0737, 0.0771, 0.3438, 1.0703, + 0.4199, 0.1035, 0.9062, 1.0000, 0.7422, 0.0762, 0.3223, 0.1289, 0.2500, + 0.8945, 0.2832], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.v_proj.weight': tensor([[ 0.0060, 0.0017, -0.0410, ..., -0.0112, -0.0168, -0.0066], + [-0.0082, -0.0021, 0.0059, ..., 0.0078, 0.0058, -0.0205], + [-0.0188, -0.0048, -0.0022, ..., -0.0126, 0.0251, -0.0205], + ..., + [-0.0201, -0.0062, -0.0140, ..., -0.0175, 0.0133, -0.0270], + [ 0.0022, 0.0143, 0.0044, ..., 0.0064, -0.0052, 0.0176], + [ 0.0381, -0.0073, 0.0008, ..., -0.0004, 0.0135, 0.0223]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.output_proj.weight': tensor([[ 0.0199, 0.0085, 0.0082, ..., 0.0055, 0.0121, -0.0139], + [-0.0063, -0.0096, -0.0077, ..., 0.0020, 0.0026, 0.0287], + [-0.0011, -0.0115, -0.0266, ..., -0.0145, -0.0106, 0.0001], + ..., + [ 0.0071, 0.0115, -0.0037, ..., 0.0310, 0.0048, -0.0071], + [ 0.0063, 0.0247, -0.0160, ..., -0.0090, 0.0093, -0.0070], + [ 0.0170, -0.0165, 0.0021, ..., -0.0017, 0.0105, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img.weight': tensor([[-0.0214, -0.0540, 0.0193, ..., 0.0079, -0.0272, 0.0322], + [-0.0206, -0.0140, -0.0107, ..., 0.0103, -0.0210, 0.0148], + [-0.0286, -0.0267, 0.0042, ..., 0.0123, -0.0043, 0.0089], + ..., + [-0.0225, 0.0337, -0.0084, ..., -0.0177, 0.0542, -0.0249], + [ 0.0033, 0.0320, -0.0417, ..., -0.0188, -0.0083, 0.0187], + [-0.0327, 0.0126, 0.0206, ..., 0.0043, -0.0122, -0.0559]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.v_img.weight': tensor([[ 0.0151, -0.0096, -0.0231, ..., -0.0214, 0.0447, 0.0136], + [-0.0171, -0.0064, 0.0077, ..., 0.0043, 0.0114, 0.0598], + [-0.0047, 0.0267, -0.0033, ..., 0.0284, -0.0074, -0.0364], + ..., + [-0.0050, -0.0033, 0.0233, ..., 0.0237, 0.0225, -0.0703], + [ 0.0226, 0.0037, -0.0091, ..., 0.0211, 0.0138, -0.0076], + [ 0.0100, 0.0236, -0.0206, ..., -0.0134, 0.0046, 0.0125]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img.weight': tensor([[ 0.0259, -0.0064, 0.0294, ..., 0.0033, 0.0270, 0.0055], + [-0.0295, -0.0270, 0.0161, ..., -0.0330, 0.0061, -0.0168], + [ 0.0201, -0.0121, 0.0124, ..., -0.0381, -0.0247, -0.0070], + ..., + [ 0.0273, -0.0182, -0.0183, ..., 0.0464, 0.0089, -0.0075], + [-0.0194, 0.0094, -0.0087, ..., -0.0231, -0.0066, -0.0466], + [ 0.0327, 0.0284, -0.0048, ..., 0.0145, 0.0140, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.16.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.16.mlp.layer1.weight': tensor([[-0.0056, 0.0089, -0.0228, ..., -0.0135, -0.0179, 0.0156], + [ 0.0054, 0.0017, 0.0068, ..., 0.0057, 0.0004, 0.0012], + [ 0.0216, 0.0034, 0.0261, ..., -0.0159, -0.0239, 0.0024], + ..., + [-0.0146, 0.0001, 0.0039, ..., 0.0090, -0.0097, -0.0153], + [ 0.0030, 0.0067, -0.0088, ..., -0.0216, 0.0106, 0.0198], + [-0.0093, 0.0018, -0.0107, ..., -0.0063, 0.0028, 0.0032]], + dtype=torch.bfloat16), 'net.blocks.16.mlp.layer2.weight': tensor([[ 0.0027, -0.0118, 0.0188, ..., -0.0087, 0.0016, -0.0083], + [ 0.0134, -0.0029, 0.0208, ..., 0.0038, -0.0123, -0.0122], + [-0.0002, 0.0024, 0.0002, ..., -0.0204, -0.0181, -0.0103], + ..., + [ 0.0085, 0.0172, 0.0172, ..., 0.0099, -0.0188, 0.0059], + [ 0.0087, 0.0037, -0.0144, ..., -0.0012, -0.0050, -0.0011], + [-0.0085, 0.0069, -0.0170, ..., 0.0153, -0.0221, 0.0025]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.1.weight': tensor([[-6.3477e-02, -6.2988e-02, -6.1035e-02, ..., 1.5736e-04, + 3.1281e-04, 2.2793e-04], + [-6.4941e-02, -6.3477e-02, -6.2012e-02, ..., 3.0899e-04, + 2.5558e-04, 2.4128e-04], + [-5.2979e-02, -5.1758e-02, -5.0781e-02, ..., 3.0518e-04, + 2.8038e-04, 4.2915e-04], + ..., + [-3.7109e-02, -3.6865e-02, -3.6621e-02, ..., 1.7262e-04, + 2.2984e-04, 2.4128e-04], + [-1.2695e-01, -1.2402e-01, -1.2109e-01, ..., -1.8024e-04, + -6.3896e-05, -2.0123e-04], + [ 5.7373e-02, 5.6885e-02, 5.5664e-02, ..., -2.3842e-04, + -3.1090e-04, -2.4223e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_self_attn.2.weight': tensor([[ 0.0048, 0.0047, 0.0047, ..., 0.0035, 0.0021, -0.0051], + [ 0.0026, 0.0028, 0.0042, ..., 0.0008, 0.0043, -0.0035], + [ 0.0022, 0.0016, 0.0052, ..., -0.0003, -0.0074, -0.0046], + ..., + [ 0.0173, 0.0168, 0.0078, ..., 0.0210, -0.0356, -0.0085], + [ 0.0193, 0.0182, 0.0070, ..., 0.0254, -0.0176, -0.0086], + [-0.0289, -0.0302, -0.0198, ..., -0.0320, -0.0120, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.1.weight': tensor([[-1.1414e-02, -1.0986e-02, -1.0681e-02, ..., -1.8692e-04, + -1.6403e-04, -2.2221e-04], + [-9.5215e-03, -1.0376e-02, -1.0376e-02, ..., -8.9645e-05, + -2.4319e-04, -1.6975e-04], + [-2.2705e-02, -2.3071e-02, -2.2705e-02, ..., -1.3447e-04, + -1.8787e-04, -1.2589e-04], + ..., + [ 4.6387e-03, 5.1575e-03, 5.2185e-03, ..., 2.3842e-04, + 9.6798e-05, 2.5368e-04], + [-1.3000e-02, -1.3000e-02, -1.3184e-02, ..., -8.9645e-05, + -2.3079e-04, -1.6117e-04], + [-1.4404e-02, -1.5259e-02, -1.4893e-02, ..., -1.2970e-04, + -1.9360e-04, -1.9455e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0027, 0.0006, 0.0007, ..., -0.0017, 0.0006, 0.0008], + [-0.0002, -0.0060, -0.0057, ..., 0.0061, -0.0060, -0.0058], + [-0.0109, -0.0008, -0.0008, ..., 0.0005, -0.0007, -0.0008], + ..., + [-0.0166, 0.0293, 0.0287, ..., -0.0291, 0.0292, 0.0288], + [ 0.0216, -0.0309, -0.0305, ..., 0.0297, -0.0302, -0.0299], + [-0.0302, 0.0381, 0.0371, ..., -0.0356, 0.0374, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.1.weight': tensor([[-1.2598e-01, -1.2305e-01, -1.2061e-01, ..., -2.1267e-04, + -1.4496e-04, -8.9169e-05], + [-6.3171e-03, -5.9509e-03, -5.4626e-03, ..., -2.2793e-04, + -1.9360e-04, -1.3542e-04], + [ 6.9336e-02, 6.8359e-02, 6.6406e-02, ..., 1.2875e-04, + 1.3828e-04, 2.4796e-04], + ..., + [ 6.2256e-02, 6.1035e-02, 5.9814e-02, ..., -1.6785e-04, + -2.3746e-04, -2.2221e-04], + [ 8.5938e-02, 8.3984e-02, 8.3008e-02, ..., 1.2207e-04, + 7.4863e-05, 1.8692e-04], + [-1.0400e-01, -1.0205e-01, -9.9121e-02, ..., -1.7166e-04, + -1.8406e-04, -1.4687e-04]], dtype=torch.bfloat16), 'net.blocks.16.adaln_modulation_mlp.2.weight': tensor([[ 0.0187, 0.0038, -0.0024, ..., -0.0001, -0.0015, 0.0117], + [ 0.0065, -0.0002, -0.0022, ..., -0.0021, -0.0022, 0.0052], + [ 0.0004, 0.0019, 0.0001, ..., -0.0002, -0.0005, 0.0030], + ..., + [ 0.0124, -0.0048, -0.0311, ..., -0.0211, -0.0282, 0.0091], + [ 0.0312, 0.0044, -0.0378, ..., -0.0250, -0.0359, 0.0305], + [-0.0127, 0.0161, 0.0518, ..., 0.0410, 0.0488, -0.0125]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_proj.weight': tensor([[-0.0116, 0.0046, 0.0132, ..., 0.0062, 0.0146, -0.0277], + [-0.0129, 0.0051, 0.0038, ..., 0.0238, -0.0132, 0.0244], + [ 0.0229, -0.0009, -0.0177, ..., 0.0258, -0.0016, -0.0073], + ..., + [-0.0018, -0.0001, -0.0014, ..., 0.0081, -0.0054, -0.0062], + [ 0.0236, -0.0055, -0.0015, ..., -0.0008, -0.0165, -0.0201], + [ 0.0063, -0.0040, -0.0110, ..., 0.0004, -0.0024, -0.0002]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm.weight': tensor([1.5000, 1.8750, 1.0625, 1.1797, 1.3594, 0.8438, 1.3125, 1.2812, 1.3828, + 1.3750, 1.4141, 1.2734, 1.5078, 1.1562, 1.3906, 1.4922, 1.3672, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.2109, 1.7734, 1.7188, 1.7500, 1.7500, + 1.3516, 1.4609, 1.3750, 1.4609, 1.3594, 1.3750, 1.4688, 1.3438, 1.3828, + 1.4531, 1.4531, 1.2734, 1.5156, 1.4297, 1.3516, 0.0889, 1.4766, 1.1641, + 1.0547, 1.1875, 1.6797, 0.8633, 1.5781, 1.2969, 1.4453, 1.3828, 1.4844, + 1.3750, 1.4297, 1.2344, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.1250, 0.7109, 0.6797, 1.6016, 0.7148, 0.4492, 1.1250, 1.4297, + 1.3359, 1.3984, 1.3828, 1.2344, 0.3711, 1.3281, 1.3438, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 0.9727, 1.3672, 1.4609, 1.2812, + 1.5156, 1.2422, 1.3359, 1.4531, 1.4766, 1.4453, 1.4219, 1.2812, 1.4766, + 1.3750, 1.3438, 1.2891, 1.4688, 1.3828, 1.1875, 1.3359, 1.3516, 0.8594, + 0.9961, 1.8750, 1.8438, 1.6094, 1.2031, 1.5234, 0.8828, 1.4531, 1.1094, + 1.4141, 1.1484, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.k_proj.weight': tensor([[ 0.0014, 0.0042, -0.0391, ..., -0.0139, 0.0044, 0.0151], + [-0.0339, 0.0041, -0.0001, ..., 0.0037, -0.0168, 0.0039], + [ 0.0442, -0.0096, -0.0002, ..., -0.0165, -0.0082, 0.0143], + ..., + [-0.0063, -0.0046, 0.0055, ..., 0.0025, 0.0056, 0.0058], + [ 0.0064, -0.0133, 0.0102, ..., -0.0031, -0.0071, 0.0034], + [-0.0023, -0.0033, -0.0036, ..., -0.0010, -0.0021, -0.0012]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm.weight': tensor([1.9375, 1.6250, 1.0078, 1.1172, 1.3281, 0.8438, 1.3047, 1.3125, 1.4062, + 1.3750, 1.4062, 1.2734, 1.5078, 1.1641, 1.3906, 1.4922, 1.3750, 1.4844, + 1.4062, 1.4141, 1.4062, 1.3594, 1.6875, 1.6328, 1.4922, 1.6172, 1.7188, + 1.2656, 1.3672, 1.3750, 1.4531, 1.3516, 1.3750, 1.4688, 1.3516, 1.3906, + 1.4531, 1.4531, 1.2656, 1.5156, 1.4297, 1.3516, 0.0884, 1.2422, 1.7031, + 1.3281, 1.3047, 1.7031, 0.8477, 1.5625, 1.2500, 1.5156, 1.3203, 1.4922, + 1.3672, 1.4375, 1.2266, 1.1719, 1.3203, 0.2910, 1.2500, 1.3984, 1.3828, + 1.5234, 1.0938, 0.8320, 0.7734, 1.5781, 0.8164, 0.4238, 1.1016, 1.4453, + 1.3125, 1.3984, 1.3984, 1.2266, 0.3750, 1.3359, 1.3516, 1.3750, 1.3906, + 1.4375, 1.2266, 1.2109, 1.2422, 1.2812, 1.6562, 1.4922, 1.4141, 1.4766, + 1.5781, 1.3281, 1.5078, 1.4609, 1.4766, 1.4375, 1.4141, 1.2891, 1.4766, + 1.3750, 1.3516, 1.2891, 1.4688, 1.3750, 1.1875, 1.3359, 1.3516, 1.1016, + 1.1484, 1.3594, 1.8125, 1.5000, 1.2188, 1.5781, 0.9102, 1.3984, 1.0859, + 1.4062, 1.1719, 1.4453, 1.4062, 1.4297, 1.4766, 1.3047, 1.1484, 1.4219, + 1.4688, 1.5156], dtype=torch.bfloat16), 'net.blocks.17.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.self_attn.v_proj.weight': tensor([[ 0.0070, 0.0112, -0.0099, ..., 0.0186, 0.0123, -0.0228], + [ 0.0170, -0.0183, 0.0119, ..., -0.0116, -0.0349, -0.0023], + [-0.0012, 0.0425, -0.0081, ..., -0.0152, -0.0114, 0.0004], + ..., + [ 0.0056, 0.0067, 0.0084, ..., 0.0123, -0.0069, 0.0094], + [ 0.0134, -0.0066, -0.0096, ..., -0.0065, 0.0225, 0.0114], + [ 0.0033, 0.0216, 0.0142, ..., -0.0179, -0.0023, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.17.self_attn.output_proj.weight': tensor([[ 0.0096, 0.0272, 0.0097, ..., 0.0188, 0.0098, -0.0095], + [ 0.0101, -0.0023, -0.0171, ..., 0.0027, 0.0082, -0.0134], + [ 0.0046, -0.0088, 0.0060, ..., -0.0106, -0.0128, 0.0060], + ..., + [-0.0012, 0.0006, 0.0038, ..., 0.0212, 0.0024, -0.0022], + [-0.0069, -0.0141, 0.0080, ..., -0.0058, 0.0106, -0.0025], + [ 0.0105, -0.0092, 0.0018, ..., -0.0061, -0.0043, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_proj.weight': tensor([[ 0.0206, -0.0045, 0.0178, ..., 0.0020, -0.0203, 0.0062], + [-0.0145, -0.0024, 0.0017, ..., 0.0139, -0.0070, -0.0131], + [ 0.0522, -0.0327, -0.0112, ..., -0.0092, -0.0020, 0.0299], + ..., + [ 0.0212, -0.0087, -0.0040, ..., -0.0034, -0.0119, -0.0017], + [ 0.0243, -0.0161, -0.0031, ..., -0.0097, 0.0146, 0.0084], + [-0.0337, -0.0081, 0.0127, ..., 0.0046, -0.0042, -0.0020]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm.weight': tensor([0.2148, 1.1484, 0.5469, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1094, + 1.1016, 0.8555, 0.1006, 0.3691, 1.0469, 1.1016, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2490, 0.0952, 1.0391, 0.9922, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1719, 0.9531, 1.1094, 0.0767, 0.1216, 0.1602, 0.3281, 0.8711, 1.0547, + 1.0938, 0.7930, 1.0547, 1.1094, 0.8594, 0.4688, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1650, 1.0312, 0.1885, 0.2773, 0.8047, 0.8633, 0.1758, + 1.1172, 0.1396, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0547, 1.0391, + 0.9180, 0.0923, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7930, 0.0569, 0.3438, 1.0547, 0.7734, 0.3789, 0.0845, + 0.6836, 1.0391, 0.0977, 0.0564, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0000, 0.2988, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1396, 1.0547, 0.9219, 0.8125, 0.6367, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_proj.weight': tensor([[ 0.0078, -0.0001, 0.0054, ..., -0.0001, 0.0032, 0.0110], + [ 0.0200, -0.0112, 0.0068, ..., -0.0133, -0.0039, -0.0105], + [ 0.0052, 0.0054, 0.0010, ..., -0.0184, -0.0223, 0.0059], + ..., + [-0.0251, -0.0022, -0.0052, ..., -0.0145, -0.0164, -0.0123], + [-0.0025, -0.0014, 0.0116, ..., 0.0060, 0.0156, -0.0016], + [ 0.0041, -0.0148, 0.0042, ..., -0.0018, 0.0082, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm.weight': tensor([0.2148, 1.1484, 0.5508, 0.1689, 1.0625, 1.1719, 1.0078, 0.7383, 0.1079, + 1.1016, 0.8555, 0.1011, 0.3691, 1.0469, 1.1094, 0.3301, 1.0938, 0.9570, + 0.8594, 0.2500, 0.0952, 1.0391, 0.9961, 0.7891, 1.1328, 1.1016, 1.0000, + 0.1729, 0.9531, 1.1094, 0.0752, 0.1216, 0.1602, 0.3281, 0.8750, 1.0547, + 1.1016, 0.7930, 1.0547, 1.1094, 0.8594, 0.4727, 1.1172, 0.0747, 0.7773, + 0.8594, 1.0859, 0.2656, 0.9453, 1.0703, 0.9219, 1.1016, 1.0938, 0.7891, + 1.0938, 0.9453, 1.0234, 1.1562, 1.0938, 1.0781, 0.3887, 1.0156, 1.1016, + 0.8750, 0.4551, 0.1641, 1.0312, 0.1895, 0.2754, 0.8086, 0.8633, 0.1758, + 1.1172, 0.1387, 0.2773, 0.2539, 0.9336, 0.1504, 1.0781, 1.0625, 1.0391, + 0.9180, 0.0908, 0.9727, 0.1738, 1.0703, 1.1016, 0.1279, 0.3398, 0.9375, + 0.5352, 0.8125, 0.7891, 0.0569, 0.3438, 1.0547, 0.7734, 0.3770, 0.0845, + 0.6836, 1.0391, 0.0981, 0.0569, 0.2891, 1.1250, 1.0156, 0.5273, 1.0000, + 0.3027, 1.1172, 0.9961, 1.1172, 1.0000, 1.0078, 0.3008, 1.0938, 0.1006, + 0.5625, 0.7656, 0.0703, 1.0703, 0.1387, 1.0547, 0.9219, 0.8125, 0.6406, + 0.9570, 1.0859], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.v_proj.weight': tensor([[-0.0192, 0.0099, -0.0013, ..., -0.0030, 0.0073, 0.0053], + [ 0.0024, -0.0043, -0.0270, ..., -0.0044, -0.0056, -0.0087], + [ 0.0223, 0.0132, -0.0204, ..., -0.0016, 0.0015, 0.0291], + ..., + [-0.0136, -0.0071, 0.0085, ..., -0.0091, 0.0072, -0.0229], + [ 0.0219, -0.0036, -0.0118, ..., -0.0056, -0.0262, -0.0061], + [-0.0075, -0.0055, -0.0164, ..., -0.0096, 0.0178, 0.0332]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.output_proj.weight': tensor([[ 0.0154, -0.0059, 0.0047, ..., -0.0020, -0.0120, 0.0245], + [ 0.0134, -0.0016, 0.0115, ..., -0.0025, 0.0016, -0.0309], + [-0.0132, 0.0067, -0.0227, ..., 0.0123, -0.0292, -0.0133], + ..., + [-0.0366, -0.0042, -0.0040, ..., 0.0099, -0.0114, -0.0021], + [-0.0238, -0.0044, -0.0204, ..., -0.0063, 0.0123, -0.0251], + [-0.0001, 0.0123, 0.0134, ..., -0.0278, -0.0082, -0.0227]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img.weight': tensor([[-0.0138, -0.0038, -0.0167, ..., -0.0034, 0.0170, 0.0108], + [ 0.0037, -0.0002, 0.0295, ..., -0.0172, 0.0009, 0.0135], + [ 0.0193, 0.0236, -0.0123, ..., 0.0049, -0.0369, 0.0239], + ..., + [ 0.0153, -0.0413, -0.0125, ..., -0.0354, 0.0059, -0.0320], + [-0.0266, -0.0062, -0.0327, ..., -0.0020, 0.0045, 0.0047], + [ 0.0298, 0.0356, -0.0192, ..., 0.0042, -0.0004, -0.0123]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.v_img.weight': tensor([[ 0.0077, 0.0059, 0.0317, ..., 0.0049, -0.0189, 0.0173], + [ 0.0164, -0.0283, -0.0454, ..., -0.0366, 0.0079, -0.0096], + [-0.0166, 0.0073, -0.0156, ..., 0.0176, -0.0148, 0.0171], + ..., + [-0.0332, -0.0050, 0.0260, ..., 0.0159, 0.0034, -0.0474], + [ 0.0038, 0.0070, -0.0074, ..., 0.0140, 0.0557, 0.0044], + [-0.0203, 0.0273, -0.0075, ..., -0.0330, -0.0016, 0.0168]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img.weight': tensor([[-0.0332, -0.0124, 0.0072, ..., 0.0039, -0.0093, -0.0277], + [ 0.0057, -0.0061, -0.0058, ..., -0.0060, 0.0140, -0.0161], + [-0.0047, 0.0168, -0.0055, ..., -0.0092, 0.0124, 0.0036], + ..., + [ 0.0398, 0.0452, 0.0251, ..., -0.0067, -0.0330, -0.0012], + [-0.0620, -0.0040, 0.0371, ..., 0.0410, 0.0041, -0.0173], + [ 0.0126, -0.0039, -0.0270, ..., -0.0043, 0.0388, -0.0103]], + dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.17.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.17.mlp.layer1.weight': tensor([[-2.4902e-02, 1.1353e-02, -1.0071e-02, ..., -1.2939e-02, + 2.3315e-02, 3.4668e-02], + [ 8.9111e-03, -5.1880e-03, 1.1673e-03, ..., 3.9673e-03, + 2.7710e-02, 2.1606e-02], + [-2.7771e-03, 9.5367e-05, -3.8757e-03, ..., -2.1515e-03, + -8.3618e-03, 8.3618e-03], + ..., + [ 8.5449e-03, -8.9722e-03, -1.4160e-02, ..., -4.2419e-03, + 4.5471e-03, -1.0864e-02], + [-3.9673e-03, -6.3782e-03, 2.3438e-02, ..., -2.4414e-03, + -1.6602e-02, 2.4567e-03], + [-1.2875e-04, -2.5482e-03, 1.3733e-02, ..., 2.4170e-02, + 1.8311e-02, -1.4404e-02]], dtype=torch.bfloat16), 'net.blocks.17.mlp.layer2.weight': tensor([[ 0.0044, 0.0143, -0.0069, ..., -0.0053, -0.0203, -0.0097], + [-0.0078, -0.0116, 0.0081, ..., 0.0374, 0.0137, 0.0075], + [ 0.0041, 0.0151, 0.0022, ..., 0.0036, 0.0121, -0.0275], + ..., + [ 0.0282, 0.0094, 0.0133, ..., 0.0035, -0.0049, -0.0175], + [-0.0082, 0.0131, -0.0093, ..., 0.0078, -0.0001, -0.0226], + [ 0.0229, -0.0288, -0.0132, ..., -0.0079, 0.0145, 0.0114]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.1.weight': tensor([[-5.5908e-02, -5.5176e-02, -5.4443e-02, ..., 3.0136e-04, + 3.0518e-04, 3.7956e-04], + [ 1.1523e-01, 1.1182e-01, 1.0840e-01, ..., -2.0981e-04, + -1.1349e-04, -1.3161e-04], + [ 1.2451e-01, 1.2207e-01, 1.1865e-01, ..., -1.4973e-04, + -7.9632e-05, -1.8215e-04], + ..., + [ 2.3438e-02, 2.3315e-02, 2.3193e-02, ..., -7.3910e-05, + -7.0095e-05, -6.1512e-05], + [ 5.6152e-02, 5.4932e-02, 5.4199e-02, ..., -1.4400e-04, + -2.0504e-04, -2.5177e-04], + [-8.3008e-02, -8.0566e-02, -7.9590e-02, ..., -2.1935e-04, + -1.8978e-04, -3.1853e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_self_attn.2.weight': tensor([[ 1.8463e-03, -1.0803e-02, -1.2573e-02, ..., -4.1504e-03, + -4.5166e-03, 1.0132e-02], + [ 2.0447e-03, 8.6670e-03, 1.0193e-02, ..., 9.2697e-04, + 4.0817e-04, -7.7515e-03], + [-2.5330e-03, -4.2200e-05, -1.9836e-03, ..., 2.4605e-04, + 4.1199e-04, 3.7689e-03], + ..., + [-7.4158e-03, 2.1362e-02, 2.1851e-02, ..., -2.2339e-02, + -1.8555e-02, 6.9275e-03], + [ 1.0376e-02, -2.3346e-03, -2.2125e-03, ..., -2.4414e-02, + -2.3071e-02, 2.9297e-03], + [-2.6001e-02, 1.7212e-02, 2.0386e-02, ..., 3.4912e-02, + 3.6377e-02, -9.4604e-03]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.1.weight': tensor([[-9.6436e-03, -9.7046e-03, -9.3384e-03, ..., 1.4687e-04, + 8.3923e-05, 1.6499e-04], + [-1.9455e-03, -1.9455e-03, -1.9455e-03, ..., -5.5313e-05, + -2.3746e-04, -1.4114e-04], + [-2.0447e-03, -2.0447e-03, -1.3885e-03, ..., 2.0027e-04, + 1.2398e-04, 1.9169e-04], + ..., + [-5.6396e-02, -5.5664e-02, -5.3955e-02, ..., 1.0538e-04, + 6.6757e-05, 2.0027e-04], + [-3.0975e-03, -2.4567e-03, -2.4414e-03, ..., -1.3542e-04, + -1.5163e-04, -8.8692e-05], + [ 1.9455e-03, -2.2736e-03, -2.7161e-03, ..., 1.5259e-04, + 1.3065e-04, 2.0123e-04]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0072, -0.0070, 0.0048, ..., -0.0010, -0.0058, 0.0070], + [-0.0025, 0.0015, -0.0040, ..., -0.0038, 0.0008, -0.0015], + [ 0.0016, -0.0025, 0.0013, ..., 0.0006, -0.0024, 0.0027], + ..., + [ 0.0342, -0.0339, -0.0186, ..., 0.0168, -0.0334, 0.0344], + [ 0.0447, -0.0439, -0.0142, ..., 0.0236, -0.0425, 0.0449], + [ 0.0378, -0.0374, -0.0408, ..., 0.0388, -0.0361, 0.0378]], + dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.1.weight': tensor([[-9.6680e-02, -9.4238e-02, -9.2773e-02, ..., -3.8385e-05, + -1.1587e-04, -1.0157e-04], + [-8.1177e-03, -8.9111e-03, -9.3994e-03, ..., 1.3351e-04, + 1.0824e-04, 1.5354e-04], + [ 8.9844e-02, 8.8379e-02, 8.6914e-02, ..., 9.6321e-05, + 2.4223e-04, 1.0729e-04], + ..., + [ 5.3223e-02, 5.1758e-02, 5.0049e-02, ..., 2.3842e-04, + 1.4877e-04, 2.4605e-04], + [ 7.4707e-02, 7.3730e-02, 7.1777e-02, ..., -5.7459e-05, + -8.1062e-05, 5.9232e-07], + [-3.8086e-02, -3.6865e-02, -3.5889e-02, ..., 2.1696e-05, + 1.3065e-04, -9.4414e-05]], dtype=torch.bfloat16), 'net.blocks.17.adaln_modulation_mlp.2.weight': tensor([[-0.0007, 0.0052, -0.0014, ..., -0.0038, 0.0038, -0.0066], + [-0.0031, 0.0007, 0.0008, ..., 0.0021, 0.0042, -0.0010], + [-0.0004, -0.0017, 0.0015, ..., 0.0028, 0.0002, 0.0019], + ..., + [ 0.0381, 0.0361, -0.0378, ..., -0.0046, -0.0305, -0.0327], + [ 0.0469, 0.0417, -0.0444, ..., -0.0042, -0.0469, -0.0364], + [-0.0591, -0.0483, 0.0554, ..., 0.0062, 0.0640, 0.0388]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_proj.weight': tensor([[ 0.0091, 0.0053, -0.0056, ..., -0.0037, -0.0042, 0.0063], + [-0.0118, -0.0179, -0.0042, ..., -0.0198, 0.0184, 0.0059], + [ 0.0220, -0.0033, -0.0011, ..., 0.0051, 0.0044, -0.0021], + ..., + [ 0.0184, 0.0206, -0.0023, ..., -0.0063, -0.0052, -0.0204], + [ 0.0012, -0.0057, -0.0019, ..., 0.0048, 0.0187, -0.0150], + [ 0.0100, -0.0042, 0.0190, ..., 0.0261, -0.0129, 0.0001]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm.weight': tensor([0.9141, 0.0186, 1.0234, 1.5391, 1.2891, 1.2109, 1.1250, 1.4375, 1.3438, + 1.3516, 1.3750, 1.3750, 1.4297, 1.4219, 1.3438, 0.8672, 1.3906, 1.4688, + 1.4141, 1.4453, 1.2891, 1.4453, 0.9961, 1.2422, 1.8594, 1.5000, 1.2344, + 1.4219, 1.5000, 1.4922, 1.3906, 1.4062, 1.3594, 1.4922, 1.4219, 1.4141, + 1.4141, 1.4609, 1.0938, 1.4688, 1.3359, 1.3438, 1.4219, 0.9453, 1.1250, + 1.6094, 1.4297, 1.1953, 1.7891, 1.1797, 1.4062, 1.4062, 0.0596, 1.0234, + 1.4531, 1.3906, 1.4062, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6719, 1.4766, 1.5859, 0.9258, 0.9141, 0.9688, 1.1562, 1.3203, + 1.3281, 1.3047, 1.3594, 1.4219, 1.3750, 1.3516, 1.4297, 1.3750, 1.3984, + 1.3672, 1.3203, 1.3516, 1.1484, 1.3828, 1.1719, 1.7969, 1.8203, 1.4688, + 1.5156, 1.4609, 1.3984, 1.5078, 1.4609, 1.3516, 1.3359, 1.3125, 1.4844, + 1.3203, 1.3203, 1.3281, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.6016, + 1.6484, 1.3516, 1.7969, 0.9414, 1.5625, 1.4297, 1.3672, 1.2969, 1.4062, + 1.4375, 1.2891, 1.4922, 1.4219, 1.2500, 1.4453, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.k_proj.weight': tensor([[ 0.0097, 0.0082, -0.0126, ..., -0.0062, -0.0026, 0.0037], + [-0.0186, 0.0073, 0.0019, ..., 0.0136, -0.0070, 0.0161], + [-0.0072, -0.0079, 0.0025, ..., -0.0042, 0.0106, -0.0065], + ..., + [ 0.0121, -0.0022, 0.0262, ..., 0.0140, -0.0127, -0.0014], + [-0.0049, -0.0131, -0.0115, ..., -0.0107, 0.0059, 0.0036], + [ 0.0145, 0.0036, -0.0125, ..., 0.0070, -0.0029, 0.0124]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm.weight': tensor([1.0859, 0.2109, 0.8477, 1.3906, 1.2109, 1.3281, 1.0781, 1.3984, 1.3438, + 1.3359, 1.3672, 1.3750, 1.4062, 1.4141, 1.3516, 0.8594, 1.3906, 1.4766, + 1.4141, 1.4453, 1.2891, 1.4375, 1.4297, 1.2344, 1.2109, 1.6719, 1.2656, + 1.3047, 1.5000, 1.4766, 1.4141, 1.3984, 1.3672, 1.4922, 1.4297, 1.4141, + 1.4062, 1.4609, 1.0938, 1.4688, 1.3359, 1.3359, 1.4219, 1.1406, 1.1484, + 1.4453, 1.3594, 1.1719, 1.6250, 1.1484, 1.3906, 1.3906, 0.0300, 1.0312, + 1.4453, 1.3906, 1.3984, 1.2109, 1.3906, 1.2188, 1.3750, 1.2969, 1.2500, + 1.2812, 1.6484, 1.4844, 1.6172, 1.2656, 1.0625, 0.8398, 1.1719, 1.2969, + 1.3203, 1.2812, 1.3516, 1.4297, 1.3828, 1.3672, 1.4219, 1.3750, 1.3984, + 1.3750, 1.3203, 1.3516, 1.1484, 1.3828, 0.8789, 1.4062, 1.4609, 1.1953, + 1.3750, 1.4375, 1.4609, 1.5078, 1.4531, 1.3203, 1.3359, 1.3281, 1.4766, + 1.3125, 1.3203, 1.3203, 1.3359, 1.3203, 1.3750, 1.3750, 1.4062, 1.5703, + 1.8203, 1.3359, 1.6562, 1.0078, 1.2500, 1.2734, 1.4375, 1.3203, 1.4062, + 1.4375, 1.3047, 1.4922, 1.4219, 1.2500, 1.4297, 1.3828, 1.4297, 1.3672, + 1.3672, 1.3750], dtype=torch.bfloat16), 'net.blocks.18.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.self_attn.v_proj.weight': tensor([[-0.0146, -0.0052, -0.0106, ..., 0.0080, 0.0035, -0.0039], + [ 0.0078, 0.0079, 0.0125, ..., -0.0117, -0.0171, 0.0251], + [-0.0026, -0.0042, 0.0087, ..., -0.0011, -0.0034, 0.0208], + ..., + [ 0.0017, 0.0111, 0.0047, ..., -0.0254, -0.0093, -0.0066], + [-0.0013, 0.0115, 0.0226, ..., 0.0108, 0.0116, -0.0238], + [ 0.0042, 0.0189, 0.0265, ..., 0.0106, 0.0107, 0.0122]], + dtype=torch.bfloat16), 'net.blocks.18.self_attn.output_proj.weight': tensor([[-0.0063, -0.0149, 0.0109, ..., -0.0060, -0.0076, -0.0080], + [-0.0153, -0.0032, -0.0183, ..., -0.0096, -0.0035, -0.0112], + [-0.0091, 0.0099, 0.0090, ..., 0.0193, -0.0145, 0.0086], + ..., + [ 0.0042, 0.0072, 0.0048, ..., -0.0130, 0.0038, -0.0030], + [ 0.0154, -0.0082, 0.0204, ..., -0.0067, 0.0068, 0.0082], + [ 0.0040, 0.0081, -0.0137, ..., -0.0087, -0.0064, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_proj.weight': tensor([[-1.1780e-02, -5.6763e-03, 6.9580e-03, ..., 1.4282e-02, + 6.3419e-05, -3.8300e-03], + [ 1.2756e-02, 7.3242e-03, -7.8735e-03, ..., 1.8616e-03, + -9.7275e-04, 9.7656e-03], + [ 1.8677e-02, -1.8188e-02, 1.1353e-02, ..., 6.9275e-03, + -1.0498e-02, -5.0659e-03], + ..., + [-2.4719e-03, 2.5513e-02, -2.0386e-02, ..., 1.9165e-02, + 7.8125e-03, 4.9133e-03], + [ 5.6763e-03, -5.2795e-03, 2.8198e-02, ..., -6.8359e-03, + 1.3184e-02, -8.3008e-03], + [ 5.6763e-03, 2.1606e-02, -1.3184e-02, ..., -1.1169e-02, + -2.6093e-03, 2.2217e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm.weight': tensor([ 1.9897e-02, 6.4453e-02, 4.3164e-01, 1.0742e-02, 8.0469e-01, + 1.0132e-02, 1.0791e-01, 2.0874e-02, 3.7109e-02, 7.8906e-01, + 5.2979e-02, 9.1406e-01, 1.2756e-02, -5.0735e-04, 3.3691e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3867e-01, -5.0783e-05, 8.1543e-02, 1.1414e-02, + 1.5747e-02, 1.6895e-01, 8.7891e-01, -3.8743e-06, 2.1118e-02, + 4.5898e-02, 2.0905e-03, 7.5684e-02, 8.0566e-02, 2.1240e-02, + 4.4678e-02, 5.7422e-01, 1.6235e-02, 1.4551e-01, 1.2360e-03, + 3.3594e-01, 2.2827e-02, 1.4062e-01, 1.7212e-02, 5.8984e-01, + 5.5176e-02, 2.9755e-04, 2.6367e-02, 3.5156e-01, 5.4443e-02, + 6.6016e-01, 8.9844e-01, 8.3008e-03, 3.4961e-01, 4.0039e-01, + 3.7956e-04, 2.2339e-02, 8.6914e-02, 1.2024e-02, 7.8125e-01, + 5.0293e-02, -2.4140e-06, 1.4453e-01, 9.3750e-02, 1.0376e-03, + 1.8359e-01, 7.3828e-01, 3.8385e-05, 1.7090e-02, 1.1536e-02, + 2.1076e-04, 1.6113e-02, 1.1377e-01, 8.7891e-03, 6.1035e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7109e-01, 2.9297e-01, + -1.8597e-05, 1.7578e-02, 3.2654e-03, 9.1406e-01, 8.0078e-01, + 3.4766e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 3.0975e-03, + 1.2085e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1475e-02, + 3.5400e-02, 7.3828e-01, 8.8672e-01, 8.3203e-01, 2.1240e-02, + 5.0964e-03, 8.1543e-02, 6.7969e-01, 2.2461e-02, 5.0964e-03, + 8.0078e-01, 1.0234e+00, 1.5747e-02, 1.4453e-01, 2.0386e-02, + 4.0894e-03, 1.0742e-02, 2.4414e-03, 1.6357e-02, 9.7656e-03, + 4.8340e-02, 5.6396e-02, 1.1658e-02, 3.7500e-01, 3.8672e-01, + 5.7373e-02, 8.0469e-01, 5.2490e-03, 2.9492e-01, 9.0332e-02, + 1.4355e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_proj.weight': tensor([[ 0.0098, 0.0021, -0.0150, ..., 0.0007, -0.0003, 0.0061], + [-0.0013, 0.0025, -0.0065, ..., 0.0053, 0.0070, -0.0021], + [-0.0090, -0.0096, -0.0117, ..., -0.0016, 0.0079, 0.0023], + ..., + [-0.0014, 0.0021, -0.0027, ..., -0.0019, 0.0054, 0.0002], + [-0.0027, -0.0020, -0.0034, ..., -0.0011, 0.0076, -0.0026], + [ 0.0151, -0.0024, 0.0025, ..., 0.0013, -0.0004, -0.0118]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm.weight': tensor([ 1.9897e-02, 6.4941e-02, 4.3164e-01, 1.0803e-02, 8.0469e-01, + 1.0437e-02, 1.0791e-01, 2.0996e-02, 3.6865e-02, 7.8906e-01, + 5.2490e-02, 9.1406e-01, 1.2634e-02, -5.0735e-04, 3.4668e-02, + 7.2754e-02, 2.2754e-01, 9.0234e-01, 7.4609e-01, 6.8359e-01, + 4.8438e-01, 1.3965e-01, -5.1260e-05, 8.2520e-02, 1.1475e-02, + 1.5625e-02, 1.6895e-01, 8.7891e-01, -4.5002e-06, 2.0752e-02, + 4.5898e-02, 2.1667e-03, 7.5684e-02, 8.0566e-02, 2.0996e-02, + 4.4678e-02, 5.7422e-01, 1.5991e-02, 1.4551e-01, 1.1749e-03, + 3.3594e-01, 2.1973e-02, 1.4062e-01, 1.7090e-02, 5.8984e-01, + 5.5176e-02, 2.8992e-04, 2.6245e-02, 3.5156e-01, 5.4688e-02, + 6.6016e-01, 8.9844e-01, 8.1177e-03, 3.4961e-01, 4.0039e-01, + 3.9482e-04, 2.2095e-02, 8.6914e-02, 1.1658e-02, 7.8125e-01, + 5.0293e-02, -2.5034e-06, 1.4453e-01, 9.4727e-02, 9.8419e-04, + 1.8457e-01, 7.3828e-01, 3.6955e-05, 1.7334e-02, 1.1658e-02, + 1.9932e-04, 1.5869e-02, 1.1377e-01, 8.6060e-03, 6.2012e-02, + 6.6016e-01, 3.8672e-01, 4.8438e-01, 8.7500e-01, 2.9297e-01, + -1.8716e-05, 1.7700e-02, 3.4332e-03, 9.1406e-01, 8.0078e-01, + 3.4961e-01, 4.8242e-01, 1.8555e-02, 2.4414e-02, 2.9602e-03, + 1.1963e-02, 8.6328e-01, 8.4766e-01, 7.1484e-01, 1.1292e-02, + 3.5156e-02, 7.3828e-01, 8.9062e-01, 8.3203e-01, 2.1362e-02, + 4.7913e-03, 8.1543e-02, 6.7578e-01, 2.2217e-02, 5.1575e-03, + 7.9688e-01, 1.0234e+00, 1.5625e-02, 1.4453e-01, 2.0874e-02, + 3.9368e-03, 1.0986e-02, 2.3956e-03, 1.6357e-02, 9.8877e-03, + 4.7363e-02, 5.6396e-02, 1.1719e-02, 3.7500e-01, 3.8672e-01, + 5.8105e-02, 8.0469e-01, 5.0964e-03, 2.9492e-01, 9.0332e-02, + 1.4160e-01, 9.5703e-02, 3.2959e-02], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.v_proj.weight': tensor([[-3.1494e-02, 2.7008e-03, 1.2756e-02, ..., -4.5776e-03, + -1.4038e-02, 9.8877e-03], + [-2.7771e-03, 1.3275e-03, 7.3910e-05, ..., 4.3869e-04, + 1.0864e-02, -2.7466e-03], + [-2.8839e-03, 4.6692e-03, -4.9438e-03, ..., 4.5471e-03, + 1.1536e-02, -1.2634e-02], + ..., + [-1.0254e-02, -9.5215e-03, -6.6223e-03, ..., -1.2817e-02, + 6.8359e-03, -1.6602e-02], + [ 9.6130e-04, -1.6968e-02, -1.4893e-02, ..., -1.2817e-02, + 7.4768e-03, -1.9165e-02], + [-7.6675e-04, 3.3936e-02, 1.1230e-02, ..., 9.3994e-03, + -3.3417e-03, 1.8677e-02]], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.output_proj.weight': tensor([[ 0.0039, 0.0021, 0.0017, ..., 0.0025, -0.0139, 0.0129], + [ 0.0014, -0.0137, 0.0089, ..., -0.0040, -0.0103, 0.0072], + [ 0.0010, 0.0154, 0.0364, ..., 0.0065, -0.0089, 0.0025], + ..., + [ 0.0017, -0.0047, -0.0063, ..., -0.0156, -0.0136, 0.0110], + [ 0.0156, 0.0200, 0.0097, ..., -0.0114, -0.0179, 0.0106], + [-0.0049, 0.0123, -0.0059, ..., -0.0098, -0.0134, -0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img.weight': tensor([[-0.0259, 0.0229, 0.0297, ..., -0.0123, -0.0127, 0.0115], + [ 0.0337, 0.0190, -0.0249, ..., -0.0242, 0.0005, 0.0245], + [-0.0449, 0.0017, 0.0188, ..., 0.0254, 0.0112, -0.0284], + ..., + [ 0.0188, -0.0288, -0.0042, ..., 0.0041, 0.0288, 0.0127], + [ 0.0330, 0.0028, 0.0062, ..., 0.0299, 0.0143, -0.0483], + [-0.0166, 0.0078, -0.0033, ..., -0.0060, 0.0109, -0.0210]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.v_img.weight': tensor([[ 0.0182, 0.0216, -0.0104, ..., 0.0349, -0.0339, 0.0352], + [ 0.0162, 0.0420, -0.0118, ..., 0.0066, 0.0040, 0.0527], + [ 0.0037, 0.0403, 0.0112, ..., 0.0417, 0.0106, 0.0087], + ..., + [ 0.0178, 0.0288, -0.0179, ..., 0.0099, -0.0095, 0.0327], + [ 0.0135, 0.0113, -0.0067, ..., 0.0093, -0.0178, -0.0254], + [-0.0530, -0.0125, 0.0281, ..., -0.0242, -0.0128, -0.0049]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img.weight': tensor([[-0.0156, -0.0089, 0.0322, ..., -0.0150, -0.0160, 0.0029], + [-0.0052, -0.0025, -0.0286, ..., 0.0053, 0.0172, 0.0106], + [ 0.0393, 0.0400, -0.0052, ..., 0.0018, 0.0083, 0.0292], + ..., + [-0.0452, -0.0305, -0.0168, ..., 0.0106, 0.0099, 0.0150], + [ 0.0014, 0.0037, 0.0079, ..., 0.0280, 0.0258, -0.0160], + [ 0.0344, -0.0079, 0.0076, ..., 0.0092, -0.0111, 0.0041]], + dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.18.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.18.mlp.layer1.weight': tensor([[-0.0195, -0.0193, 0.0022, ..., 0.0013, -0.0413, -0.0036], + [ 0.0071, 0.0049, -0.0061, ..., 0.0051, 0.0205, 0.0060], + [ 0.0167, 0.0013, 0.0054, ..., -0.0239, 0.0064, 0.0304], + ..., + [-0.0156, -0.0107, -0.0065, ..., 0.0278, 0.0162, -0.0055], + [ 0.0134, -0.0112, 0.0112, ..., -0.0029, -0.0211, -0.0022], + [ 0.0061, -0.0104, 0.0069, ..., -0.0186, -0.0043, 0.0113]], + dtype=torch.bfloat16), 'net.blocks.18.mlp.layer2.weight': tensor([[ 0.0123, -0.0025, -0.0011, ..., -0.0021, -0.0060, 0.0038], + [-0.0165, 0.0010, -0.0083, ..., -0.0030, 0.0178, 0.0099], + [ 0.0023, -0.0278, 0.0104, ..., -0.0026, -0.0056, -0.0019], + ..., + [ 0.0099, 0.0069, -0.0173, ..., -0.0305, -0.0164, 0.0051], + [-0.0024, 0.0014, -0.0245, ..., -0.0225, -0.0060, 0.0038], + [-0.0165, -0.0018, -0.0049, ..., -0.0133, 0.0123, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.1.weight': tensor([[-6.2012e-02, -6.0791e-02, -6.0059e-02, ..., 3.4332e-04, + 2.5368e-04, 2.4033e-04], + [ 2.1851e-02, 2.0142e-02, 1.8921e-02, ..., -2.8419e-04, + -1.2207e-04, -1.7071e-04], + [ 9.8145e-02, 9.6191e-02, 9.3750e-02, ..., -1.9169e-04, + -1.3542e-04, -1.2398e-04], + ..., + [ 6.1768e-02, 6.0059e-02, 5.9326e-02, ..., -4.1580e-04, + -3.8910e-04, -3.6812e-04], + [ 1.4893e-02, 1.4587e-02, 1.3733e-02, ..., 1.5736e-04, + 2.0504e-04, 2.0599e-04], + [ 9.3750e-02, 9.1309e-02, 8.9844e-02, ..., -2.1172e-04, + -8.5831e-05, -6.1035e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_self_attn.2.weight': tensor([[ 0.0056, 0.0047, -0.0051, ..., -0.0058, -0.0018, -0.0045], + [-0.0043, -0.0007, -0.0011, ..., 0.0049, -0.0055, -0.0014], + [-0.0038, -0.0038, 0.0005, ..., 0.0038, -0.0048, 0.0015], + ..., + [ 0.0193, 0.0247, 0.0023, ..., -0.0162, -0.0141, 0.0048], + [ 0.0238, 0.0254, -0.0060, ..., -0.0239, -0.0043, -0.0017], + [-0.0352, -0.0342, 0.0055, ..., 0.0339, 0.0012, 0.0058]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.1.weight': tensor([[ 1.3245e-02, 1.3306e-02, 1.3550e-02, ..., 2.1362e-04, + 3.2997e-04, 9.8705e-05], + [-8.7280e-03, -8.5449e-03, -8.7891e-03, ..., 1.2016e-04, + 8.0585e-05, 8.8692e-05], + [ 2.0874e-02, 2.0996e-02, 2.0996e-02, ..., 2.0599e-04, + 2.1744e-04, 3.2616e-04], + ..., + [-1.5869e-02, -1.6602e-02, -1.6113e-02, ..., -1.2684e-04, + -1.4114e-04, -1.9932e-04], + [ 1.7090e-02, 1.7334e-02, 1.6846e-02, ..., 1.7071e-04, + 2.3365e-04, 1.0204e-04], + [-9.4604e-03, -1.0132e-02, -1.0010e-02, ..., 1.9360e-04, + 5.5790e-05, 8.9645e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0110, 0.0069, 0.0137, ..., -0.0139, 0.0137, 0.0059], + [ 0.0012, 0.0030, 0.0010, ..., -0.0004, 0.0004, 0.0056], + [-0.0040, -0.0024, -0.0036, ..., 0.0036, -0.0038, -0.0020], + ..., + [ 0.0254, -0.0199, 0.0289, ..., -0.0298, 0.0300, -0.0172], + [ 0.0195, -0.0260, 0.0232, ..., -0.0238, 0.0238, -0.0226], + [-0.0303, 0.0374, -0.0332, ..., 0.0339, -0.0334, 0.0299]], + dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.1.weight': tensor([[-1.2988e-01, -1.2695e-01, -1.2451e-01, ..., -7.6771e-05, + -7.9155e-05, -1.0967e-04], + [ 1.7212e-02, 1.7212e-02, 1.7944e-02, ..., -1.7929e-04, + -1.3638e-04, -8.9169e-05], + [ 6.5613e-03, 6.4697e-03, 6.5918e-03, ..., -1.0300e-04, + -1.0872e-04, -2.6464e-05], + ..., + [ 7.9590e-02, 7.8125e-02, 7.7148e-02, ..., -3.1948e-05, + 9.0003e-06, -1.0920e-04], + [ 1.5137e-02, 1.4465e-02, 1.3977e-02, ..., 1.1396e-04, + 5.1498e-05, -8.9645e-05], + [ 9.2285e-02, 9.0332e-02, 8.9355e-02, ..., 7.5817e-05, + 7.3388e-07, 9.7275e-05]], dtype=torch.bfloat16), 'net.blocks.18.adaln_modulation_mlp.2.weight': tensor([[ 2.7008e-03, -3.3951e-04, -4.3488e-04, ..., -3.2959e-03, + -4.4250e-04, -1.3351e-03], + [-5.7697e-05, 6.8665e-04, 7.3624e-04, ..., 2.6703e-04, + -6.8283e-04, 6.4087e-04], + [ 1.1139e-03, 2.2278e-03, 2.6550e-03, ..., -2.5630e-05, + -1.8463e-03, 1.2741e-03], + ..., + [ 4.7607e-02, -3.5156e-02, -3.3447e-02, ..., -4.9561e-02, + 3.0518e-02, -4.5654e-02], + [ 4.6631e-02, -3.9795e-02, -3.8086e-02, ..., -4.9805e-02, + 3.7842e-02, -4.6631e-02], + [-5.4932e-02, 5.1514e-02, 4.9805e-02, ..., 6.0303e-02, + -4.8340e-02, 5.7373e-02]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_proj.weight': tensor([[ 0.0029, 0.0114, -0.0044, ..., -0.0160, -0.0056, 0.0035], + [ 0.0006, 0.0078, 0.0123, ..., 0.0125, -0.0060, 0.0135], + [-0.0126, -0.0150, -0.0092, ..., -0.0186, 0.0114, 0.0089], + ..., + [ 0.0184, -0.0026, 0.0231, ..., 0.0107, 0.0029, -0.0081], + [-0.0255, -0.0127, -0.0008, ..., -0.0067, 0.0107, 0.0114], + [-0.0027, 0.0136, 0.0085, ..., 0.0179, -0.0029, 0.0369]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm.weight': tensor([0.4785, 1.0859, 0.2520, 1.0000, 0.8828, 1.0547, 1.2266, 1.2734, 1.1953, + 1.2734, 1.2344, 1.3281, 1.2812, 1.2656, 1.2969, 1.2422, 1.3203, 0.7695, + 1.1172, 1.2266, 1.3438, 1.3984, 1.8203, 1.7500, 1.6094, 1.3438, 1.1484, + 0.8164, 1.5000, 1.5078, 1.2578, 1.3594, 1.2812, 1.3516, 1.3125, 1.2734, + 1.3359, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 0.7695, 1.6328, + 1.5078, 1.1953, 1.1797, 1.2422, 0.8086, 1.3438, 1.4219, 1.3516, 1.3281, + 1.2812, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.8203, 1.5000, 1.3359, 1.2734, 1.5078, 1.0625, 1.1484, 1.2969, + 1.2812, 1.3438, 1.1953, 1.2734, 1.2344, 1.3516, 1.3047, 1.3047, 1.3906, + 1.3125, 1.3750, 1.3672, 1.1797, 1.2500, 0.6523, 1.0938, 1.5312, 1.7578, + 1.7734, 1.0078, 1.4531, 1.3047, 1.3281, 1.4297, 1.0781, 1.3047, 1.0547, + 1.2031, 1.2734, 1.1328, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.8125, + 1.1484, 1.3516, 1.8281, 1.0234, 1.8672, 0.6406, 1.2812, 1.4297, 1.2656, + 1.2422, 1.3438, 0.3789, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.2031, 1.3281], dtype=torch.bfloat16), 'net.blocks.19.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.k_proj.weight': tensor([[ 3.3875e-03, -8.5449e-03, -2.3804e-02, ..., -6.7444e-03, + -5.0964e-03, 4.5776e-03], + [ 4.9438e-03, 4.0588e-03, -6.6223e-03, ..., -8.5449e-03, + 6.1035e-03, 6.0730e-03], + [ 1.1292e-02, -8.5831e-05, 3.2959e-03, ..., 1.2756e-02, + -1.9531e-03, 7.6599e-03], + ..., + [ 2.3804e-02, -1.0071e-02, 2.6123e-02, ..., -2.9907e-03, + -5.0964e-03, 1.6235e-02], + [-1.7700e-02, -1.2146e-02, 2.5940e-03, ..., 7.1106e-03, + 4.1580e-04, 3.6926e-03], + [-1.5259e-03, 6.3477e-03, 8.4839e-03, ..., 1.0986e-03, + -7.0801e-03, -4.1809e-03]], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm.weight': tensor([0.8984, 0.7539, 0.8398, 0.5938, 1.0234, 1.0547, 1.2109, 1.2812, 1.1953, + 1.3047, 1.2422, 1.3203, 1.2812, 1.2578, 1.2969, 1.2344, 1.3281, 0.7695, + 1.1250, 1.2266, 1.3438, 1.3906, 1.4141, 1.6328, 1.3516, 1.3906, 1.0938, + 0.8438, 1.4922, 1.4688, 1.2578, 1.3750, 1.2734, 1.3438, 1.3281, 1.2734, + 1.3281, 1.0859, 1.2344, 1.2266, 1.4375, 0.5273, 1.1953, 1.0547, 1.5859, + 1.3125, 1.3828, 1.1016, 1.2578, 0.7930, 1.4297, 1.3594, 1.3359, 1.3125, + 1.2734, 1.3359, 1.3672, 1.3281, 1.2578, 1.2266, 1.2656, 1.1094, 1.2656, + 1.0938, 1.7109, 1.6016, 1.1250, 1.4453, 1.4297, 1.0391, 1.1250, 1.3672, + 1.2578, 1.3438, 1.2031, 1.2812, 1.2344, 1.3594, 1.3047, 1.3125, 1.3984, + 1.3047, 1.3750, 1.3594, 1.1797, 1.2500, 1.2422, 1.1562, 1.5312, 1.5938, + 1.7344, 0.9375, 1.4609, 1.3281, 1.3203, 1.4141, 1.1016, 1.2969, 1.0547, + 1.2031, 1.2656, 1.1250, 1.2422, 1.1016, 1.2969, 0.9922, 1.3281, 1.7344, + 1.2500, 1.5078, 1.7969, 1.0625, 1.6953, 0.6758, 1.2266, 1.5547, 1.2812, + 1.2422, 1.3359, 0.3848, 1.3047, 1.3750, 1.2656, 1.2109, 1.2656, 1.2578, + 1.1953, 1.3125], dtype=torch.bfloat16), 'net.blocks.19.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.self_attn.v_proj.weight': tensor([[ 0.0085, 0.0164, -0.0172, ..., 0.0012, -0.0050, 0.0165], + [ 0.0043, -0.0032, 0.0228, ..., 0.0140, 0.0007, 0.0027], + [-0.0042, -0.0479, 0.0125, ..., 0.0165, 0.0135, -0.0039], + ..., + [ 0.0040, -0.0293, -0.0103, ..., 0.0052, 0.0080, -0.0309], + [-0.0077, 0.0112, 0.0193, ..., 0.0228, 0.0199, -0.0171], + [ 0.0105, -0.0049, 0.0070, ..., -0.0017, 0.0019, -0.0140]], + dtype=torch.bfloat16), 'net.blocks.19.self_attn.output_proj.weight': tensor([[-0.0047, -0.0005, 0.0078, ..., 0.0016, -0.0095, 0.0129], + [-0.0172, 0.0071, 0.0053, ..., 0.0143, -0.0066, 0.0066], + [-0.0092, -0.0038, 0.0044, ..., -0.0087, 0.0065, 0.0297], + ..., + [-0.0089, 0.0211, -0.0112, ..., -0.0056, 0.0243, 0.0039], + [-0.0073, 0.0073, 0.0042, ..., -0.0011, -0.0045, 0.0099], + [ 0.0037, 0.0023, 0.0149, ..., 0.0045, 0.0027, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_proj.weight': tensor([[-0.0103, -0.0101, 0.0037, ..., 0.0084, -0.0143, -0.0026], + [ 0.0118, -0.0040, 0.0004, ..., -0.0322, 0.0212, 0.0172], + [-0.0120, 0.0046, -0.0157, ..., 0.0317, 0.0074, 0.0217], + ..., + [ 0.0064, -0.0093, -0.0108, ..., 0.0100, -0.0011, -0.0066], + [-0.0325, 0.0141, -0.0045, ..., -0.0082, -0.0176, -0.0127], + [ 0.0070, 0.0035, -0.0078, ..., -0.0057, 0.0234, -0.0027]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm.weight': tensor([ 3.7500e-01, 9.6436e-03, 2.0898e-01, 8.3984e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 4.7302e-03, + 4.4531e-01, 1.0315e-02, 8.1641e-01, 5.0049e-02, 1.6724e-02, + 2.4780e-02, 2.3145e-01, 1.9531e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0508e-02, 6.3281e-01, 1.8555e-02, + 9.5825e-03, 5.5237e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.8125e-03, 3.8452e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4658e-02, 1.5076e-02, -1.5974e-05, 6.1951e-03, + 2.5879e-02, 1.1963e-02, 2.5781e-01, 3.8330e-02, 6.0547e-01, + 9.3994e-03, 2.1973e-02, -2.9206e-05, 3.2812e-01, 1.4648e-01, + 1.7944e-02, 2.6978e-02, 4.2480e-02, 9.3994e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3672e-02, 2.0508e-02, 2.0508e-02, 8.0859e-01, 1.1047e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.1787e-03, + 1.2268e-02, 2.6093e-03, 1.3574e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, -2.4872e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 5.6763e-03, 8.9844e-01, 7.9346e-03, 6.1279e-02, 3.9673e-03, + 1.1230e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1738e-03, + 4.8096e-02, 1.2573e-02, 1.1158e-04, 3.0469e-01, 6.2012e-02, + 5.3516e-01, 5.1562e-01, 2.4414e-01, 2.8076e-02, 3.4424e-02, + 3.0136e-04, 1.3965e-01, 6.5918e-02, 1.8799e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 2.4080e-05, 8.3984e-02, 8.1641e-01, + 1.2793e-01, 6.3281e-01, 1.2268e-02, 7.5378e-03, 7.2266e-01, + 4.6875e-02, 6.3479e-06, 2.2705e-02, 3.3203e-02, 4.4189e-02, + 1.3733e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 1.5163e-04, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_proj.weight': tensor([[ 0.0058, -0.0027, -0.0028, ..., 0.0029, 0.0043, -0.0028], + [ 0.0067, 0.0030, 0.0009, ..., -0.0005, 0.0025, -0.0015], + [-0.0052, 0.0042, -0.0093, ..., 0.0002, 0.0041, 0.0009], + ..., + [-0.0046, 0.0109, -0.0016, ..., 0.0058, -0.0033, -0.0186], + [ 0.0018, -0.0020, -0.0125, ..., -0.0317, 0.0134, 0.0072], + [-0.0216, -0.0109, 0.0022, ..., 0.0028, 0.0059, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm.weight': tensor([ 3.7500e-01, 9.4604e-03, 2.0898e-01, 8.3496e-02, 1.3000e-02, + 8.4766e-01, 7.8125e-01, 2.5586e-01, 1.8188e-02, 5.0354e-03, + 4.4531e-01, 1.0071e-02, 8.1641e-01, 5.0537e-02, 1.6968e-02, + 2.3926e-02, 2.3145e-01, 1.9653e-02, 9.6436e-03, 5.1172e-01, + 3.4570e-01, 1.1963e-01, 2.0264e-02, 6.3281e-01, 1.8677e-02, + 9.5825e-03, 5.5542e-03, 6.7969e-01, 6.6895e-02, 2.9419e-02, + 7.9346e-03, 3.7994e-03, 8.5938e-01, 9.2773e-03, 8.6719e-01, + 1.3489e-02, 2.4414e-02, 1.4709e-02, -1.6093e-05, 6.1646e-03, + 2.5391e-02, 1.1475e-02, 2.5781e-01, 3.8574e-02, 6.0547e-01, + 9.3384e-03, 2.2095e-02, -2.9206e-05, 3.2812e-01, 1.4453e-01, + 1.7700e-02, 2.7466e-02, 4.2480e-02, 9.3384e-03, 7.9297e-01, + 6.1719e-01, 6.4062e-01, 1.4465e-02, 1.8799e-02, 1.7383e-01, + 1.3428e-02, 2.0874e-02, 2.0752e-02, 8.0859e-01, 1.1230e-02, + 1.0742e-01, 3.0859e-01, 8.0078e-01, 7.2266e-01, 8.4839e-03, + 1.2573e-02, 2.7161e-03, 1.3477e-01, 7.3047e-01, 2.5391e-01, + 1.3794e-02, 2.7161e-03, 9.2163e-03, 7.2656e-01, 9.5215e-02, + 6.0425e-03, 8.9844e-01, 7.7820e-03, 6.1279e-02, 3.9368e-03, + 1.1108e-02, 6.9922e-01, 3.4570e-01, 7.5391e-01, 3.1128e-03, + 4.8096e-02, 1.2390e-02, 1.1969e-04, 3.0469e-01, 6.3965e-02, + 5.3516e-01, 5.1562e-01, 2.4512e-01, 2.8198e-02, 3.4424e-02, + 3.3760e-04, 1.3770e-01, 6.5918e-02, 1.9653e-02, 3.7354e-02, + 2.1484e-01, 8.6914e-02, 4.6015e-05, 8.2031e-02, 8.1641e-01, + 1.2695e-01, 6.3281e-01, 1.2451e-02, 7.5073e-03, 7.2266e-01, + 4.6631e-02, 7.2718e-06, 2.2705e-02, 3.3203e-02, 4.3945e-02, + 1.3428e-02, 1.2451e-02, 5.5078e-01, 9.7656e-03, 9.5367e-05, + 4.5508e-01, 3.3398e-01, 5.7031e-01], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.v_proj.weight': tensor([[ 0.0035, -0.0160, -0.0100, ..., -0.0034, -0.0106, -0.0148], + [-0.0077, 0.0126, -0.0028, ..., -0.0011, 0.0079, 0.0176], + [-0.0099, -0.0156, -0.0193, ..., -0.0076, -0.0112, -0.0201], + ..., + [-0.0063, -0.0022, 0.0107, ..., 0.0078, 0.0177, 0.0160], + [ 0.0125, 0.0112, 0.0031, ..., 0.0075, -0.0177, -0.0084], + [-0.0129, -0.0114, 0.0053, ..., -0.0007, 0.0208, -0.0188]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.output_proj.weight': tensor([[ 0.0023, 0.0071, 0.0060, ..., 0.0124, 0.0206, -0.0096], + [ 0.0059, 0.0072, 0.0013, ..., 0.0099, -0.0077, 0.0021], + [ 0.0110, -0.0132, -0.0108, ..., 0.0095, 0.0112, 0.0031], + ..., + [ 0.0059, 0.0018, 0.0052, ..., -0.0040, 0.0175, -0.0203], + [-0.0089, 0.0161, 0.0089, ..., -0.0007, 0.0200, -0.0165], + [-0.0065, 0.0092, -0.0003, ..., 0.0045, 0.0058, 0.0016]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img.weight': tensor([[ 0.0280, -0.0075, -0.0038, ..., -0.0220, -0.0145, -0.0047], + [-0.0166, 0.0189, -0.0050, ..., 0.0096, 0.0016, 0.0364], + [-0.0079, 0.0179, -0.0452, ..., 0.0043, -0.0347, 0.0549], + ..., + [ 0.0159, 0.0439, -0.0070, ..., -0.0049, 0.0403, -0.0162], + [-0.0128, -0.0298, 0.0245, ..., -0.0272, 0.0403, -0.0188], + [-0.0148, -0.0083, 0.0065, ..., 0.0208, 0.0123, 0.0160]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.v_img.weight': tensor([[ 0.0125, 0.0289, 0.0125, ..., 0.0135, 0.0065, -0.0105], + [ 0.0188, -0.0197, -0.0054, ..., 0.0284, -0.0342, -0.0066], + [-0.0525, -0.0214, 0.0153, ..., -0.0270, 0.0083, 0.0254], + ..., + [-0.0055, 0.0085, 0.0281, ..., 0.0129, -0.0035, -0.0065], + [ 0.0134, 0.0276, -0.0212, ..., 0.0106, -0.0046, 0.0033], + [-0.0273, -0.0747, 0.0208, ..., -0.0047, 0.0046, -0.0193]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img.weight': tensor([[ 0.0056, 0.0148, 0.0334, ..., 0.0205, -0.0273, -0.0020], + [-0.0165, 0.0110, -0.0135, ..., 0.0106, -0.0067, 0.0219], + [ 0.0046, 0.0227, -0.0055, ..., 0.0126, -0.0061, 0.0151], + ..., + [-0.0347, -0.0148, 0.0062, ..., 0.0040, 0.0030, -0.0093], + [-0.0073, 0.0398, -0.0344, ..., 0.0245, 0.0635, -0.0118], + [-0.0039, -0.0500, -0.0232, ..., 0.0320, 0.0037, -0.0038]], + dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.19.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.19.mlp.layer1.weight': tensor([[-0.0124, 0.0075, 0.0019, ..., -0.0038, -0.0022, -0.0023], + [ 0.0143, 0.0128, -0.0074, ..., -0.0081, 0.0118, -0.0002], + [-0.0078, 0.0041, -0.0170, ..., 0.0096, 0.0260, -0.0039], + ..., + [-0.0043, 0.0234, 0.0044, ..., 0.0079, -0.0157, 0.0044], + [-0.0170, -0.0025, -0.0076, ..., 0.0041, 0.0093, 0.0027], + [ 0.0148, 0.0052, -0.0109, ..., -0.0079, -0.0334, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.19.mlp.layer2.weight': tensor([[ 0.0025, -0.0156, -0.0156, ..., -0.0054, -0.0220, 0.0046], + [-0.0220, -0.0048, -0.0015, ..., -0.0272, 0.0065, -0.0103], + [-0.0020, -0.0027, -0.0068, ..., 0.0107, -0.0063, 0.0063], + ..., + [-0.0074, -0.0003, 0.0052, ..., 0.0062, -0.0042, 0.0050], + [ 0.0088, -0.0052, 0.0210, ..., -0.0017, -0.0045, 0.0121], + [-0.0098, 0.0055, -0.0071, ..., 0.0260, 0.0108, 0.0273]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.1.weight': tensor([[-0.0366, -0.0359, -0.0359, ..., -0.0002, -0.0001, -0.0001], + [ 0.0179, 0.0176, 0.0172, ..., -0.0003, -0.0002, -0.0003], + [ 0.0742, 0.0728, 0.0708, ..., -0.0002, -0.0003, -0.0003], + ..., + [-0.0464, -0.0454, -0.0452, ..., -0.0002, -0.0001, -0.0001], + [-0.0693, -0.0679, -0.0669, ..., 0.0003, 0.0003, 0.0002], + [ 0.0571, 0.0559, 0.0552, ..., -0.0002, -0.0004, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_self_attn.2.weight': tensor([[ 0.0046, -0.0128, -0.0125, ..., 0.0052, 0.0089, -0.0143], + [-0.0017, -0.0013, -0.0006, ..., -0.0016, -0.0002, -0.0009], + [ 0.0008, 0.0055, 0.0038, ..., 0.0009, -0.0003, 0.0044], + ..., + [ 0.0247, -0.0159, -0.0299, ..., 0.0264, 0.0315, -0.0289], + [ 0.0242, -0.0078, -0.0271, ..., 0.0248, 0.0273, -0.0240], + [-0.0327, 0.0175, 0.0327, ..., -0.0337, -0.0349, 0.0308]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.1.weight': tensor([[-1.9775e-02, -1.9287e-02, -1.8677e-02, ..., 1.6594e-04, + 7.1526e-05, 5.3406e-05], + [ 2.5391e-02, 2.4902e-02, 2.4170e-02, ..., -1.0061e-04, + -1.8692e-04, -1.5354e-04], + [-5.6152e-02, -5.4688e-02, -5.2979e-02, ..., 1.5736e-04, + 1.5926e-04, 1.8406e-04], + ..., + [-2.8564e-02, -2.8198e-02, -2.7588e-02, ..., 1.6117e-04, + 4.6968e-05, 6.6280e-05], + [-6.4453e-02, -6.2500e-02, -6.1768e-02, ..., -8.9169e-05, + -2.6703e-04, -1.4877e-04], + [ 1.2085e-02, 1.1414e-02, 1.1902e-02, ..., -5.1975e-05, + -1.2589e-04, -1.8954e-05]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_cross_attn.2.weight': tensor([[-0.0044, 0.0016, 0.0016, ..., -0.0042, 0.0037, 0.0042], + [ 0.0026, 0.0056, -0.0018, ..., 0.0023, 0.0059, -0.0027], + [-0.0024, 0.0033, 0.0012, ..., -0.0010, 0.0014, 0.0023], + ..., + [-0.0280, -0.0179, 0.0026, ..., -0.0277, -0.0312, 0.0280], + [-0.0267, -0.0181, 0.0098, ..., -0.0270, -0.0297, 0.0258], + [ 0.0261, 0.0248, -0.0227, ..., 0.0256, 0.0214, -0.0259]], + dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.1.weight': tensor([[ 8.7402e-02, 8.5938e-02, 8.4473e-02, ..., 1.0848e-05, + 7.3910e-05, 8.0585e-05], + [-3.9307e-02, -3.8086e-02, -3.8330e-02, ..., -1.8406e-04, + -2.0790e-04, -2.6894e-04], + [-6.4941e-02, -6.3965e-02, -6.2988e-02, ..., -1.4114e-04, + -1.1492e-04, -9.1076e-05], + ..., + [-7.0312e-02, -6.8359e-02, -6.8359e-02, ..., 3.9041e-06, + 4.6492e-05, -1.6332e-05], + [-8.1055e-02, -7.9590e-02, -7.8613e-02, ..., -1.2064e-04, + -2.1935e-04, -5.8174e-05], + [ 9.0820e-02, 8.9844e-02, 8.7402e-02, ..., 3.2997e-04, + 2.5177e-04, 1.8692e-04]], dtype=torch.bfloat16), 'net.blocks.19.adaln_modulation_mlp.2.weight': tensor([[ 0.0008, -0.0011, -0.0010, ..., -0.0012, -0.0007, -0.0067], + [ 0.0009, -0.0015, -0.0011, ..., -0.0012, -0.0009, 0.0042], + [-0.0002, 0.0004, 0.0002, ..., 0.0004, 0.0003, 0.0023], + ..., + [-0.0471, 0.0403, 0.0435, ..., 0.0437, 0.0454, -0.0067], + [-0.0457, 0.0444, 0.0452, ..., 0.0457, 0.0452, -0.0199], + [ 0.0571, -0.0505, -0.0535, ..., -0.0527, -0.0562, 0.0006]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_proj.weight': tensor([[-0.0125, -0.0267, -0.0044, ..., -0.0148, -0.0040, -0.0021], + [-0.0040, 0.0139, 0.0036, ..., 0.0074, 0.0132, 0.0040], + [ 0.0210, 0.0132, -0.0192, ..., 0.0045, 0.0273, -0.0204], + ..., + [-0.0112, -0.0198, 0.0084, ..., 0.0281, -0.0145, 0.0093], + [-0.0106, 0.0137, 0.0113, ..., 0.0122, -0.0086, 0.0076], + [ 0.0047, 0.0097, -0.0023, ..., -0.0079, -0.0039, 0.0195]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm.weight': tensor([1.5000, 0.8828, 1.0234, 1.1641, 1.3828, 1.0859, 1.1719, 1.2344, 1.2344, + 1.3047, 1.4609, 1.3047, 1.4219, 1.3359, 1.3438, 1.3750, 1.3438, 1.4141, + 1.3359, 1.4062, 1.3828, 1.1641, 1.4766, 1.0859, 1.6797, 1.0781, 1.0234, + 1.5156, 1.3906, 1.4688, 1.4531, 1.3594, 1.4844, 1.3984, 1.2578, 1.3281, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.7305, 1.5000, + 1.1875, 1.5781, 0.8164, 1.4141, 1.2578, 1.3516, 1.3984, 1.3906, 1.4766, + 1.4062, 1.3984, 1.4297, 1.4219, 1.4219, 1.1484, 1.4062, 1.4219, 1.4609, + 1.2500, 0.9414, 1.6484, 1.1016, 1.4297, 0.8242, 0.4102, 0.6016, 1.2422, + 1.3281, 1.2812, 1.3359, 1.3984, 1.3125, 1.4219, 1.3438, 1.4531, 1.4609, + 1.1953, 1.3828, 1.3984, 1.1953, 1.1797, 1.0781, 1.3047, 1.2344, 1.3516, + 0.8945, 1.4844, 1.4453, 1.3906, 1.4375, 1.3906, 1.3906, 1.5078, 1.3984, + 1.1953, 1.1641, 1.4297, 1.3750, 1.4922, 1.3906, 1.4453, 1.4062, 1.1172, + 1.3203, 0.9883, 1.1641, 0.5859, 1.7188, 1.2266, 1.3438, 1.3984, 1.3672, + 1.3672, 1.3672, 1.3750, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.k_proj.weight': tensor([[ 0.0265, 0.0188, 0.0051, ..., 0.0090, 0.0056, -0.0035], + [-0.0081, -0.0054, -0.0080, ..., -0.0044, -0.0062, 0.0040], + [-0.0123, -0.0156, 0.0042, ..., -0.0134, -0.0137, 0.0047], + ..., + [ 0.0057, 0.0203, -0.0055, ..., 0.0125, -0.0039, 0.0145], + [-0.0195, 0.0204, 0.0054, ..., 0.0124, 0.0037, -0.0016], + [-0.0103, 0.0064, -0.0153, ..., 0.0060, -0.0086, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm.weight': tensor([1.3203, 0.9453, 1.0781, 0.9727, 1.3438, 1.1094, 1.1797, 1.2031, 1.2266, + 1.3047, 1.4219, 1.2969, 1.4062, 1.3359, 1.3516, 1.3672, 1.3359, 1.4141, + 1.3359, 1.4062, 1.3750, 1.1641, 1.4766, 0.9766, 1.6562, 0.7422, 0.9688, + 1.4844, 1.4062, 1.4297, 1.4531, 1.3594, 1.4688, 1.3984, 1.2500, 1.3203, + 1.4766, 1.4062, 1.2031, 1.3984, 1.3984, 1.4844, 1.4141, 0.9648, 1.1953, + 1.2578, 1.5078, 0.8164, 1.2109, 1.1016, 1.3516, 1.4062, 1.3906, 1.4531, + 1.4062, 1.3906, 1.4297, 1.4219, 1.4219, 1.1562, 1.4062, 1.4219, 1.4609, + 1.2500, 1.1953, 1.3828, 1.0703, 1.4531, 0.9688, 0.3672, 0.5508, 1.2188, + 1.3281, 1.2891, 1.3594, 1.3984, 1.3125, 1.4062, 1.3359, 1.4609, 1.4609, + 1.1875, 1.3906, 1.3984, 1.1875, 1.1797, 1.2266, 1.3906, 1.4922, 1.4297, + 0.9492, 1.4453, 1.4531, 1.3984, 1.4375, 1.3750, 1.4062, 1.5000, 1.3906, + 1.1953, 1.1719, 1.4375, 1.3750, 1.4922, 1.3828, 1.4531, 1.4062, 0.6836, + 1.3594, 0.7539, 1.4375, 0.6133, 1.5703, 1.2969, 1.3438, 1.3672, 1.3594, + 1.3828, 1.3672, 1.3828, 1.4766, 1.3203, 1.4219, 1.3672, 1.1484, 1.0547, + 1.4297, 1.0312], dtype=torch.bfloat16), 'net.blocks.20.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0149, -0.0015, ..., 0.0075, -0.0074, -0.0135], + [-0.0024, 0.0107, -0.0068, ..., -0.0043, 0.0016, -0.0050], + [ 0.0021, -0.0157, 0.0225, ..., -0.0310, 0.0014, 0.0042], + ..., + [ 0.0078, -0.0123, 0.0356, ..., -0.0269, 0.0090, -0.0156], + [-0.0003, -0.0214, -0.0038, ..., 0.0425, -0.0051, 0.0027], + [-0.0073, 0.0153, -0.0041, ..., -0.0032, 0.0025, 0.0045]], + dtype=torch.bfloat16), 'net.blocks.20.self_attn.output_proj.weight': tensor([[-0.0098, -0.0005, 0.0103, ..., -0.0005, -0.0078, -0.0148], + [-0.0249, -0.0060, 0.0143, ..., -0.0139, -0.0012, -0.0269], + [ 0.0069, 0.0054, 0.0228, ..., 0.0359, 0.0079, 0.0016], + ..., + [ 0.0096, -0.0208, -0.0354, ..., -0.0091, 0.0228, -0.0159], + [-0.0112, -0.0161, 0.0093, ..., 0.0095, -0.0091, 0.0156], + [-0.0015, -0.0123, 0.0070, ..., 0.0110, -0.0019, -0.0095]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_proj.weight': tensor([[-2.2461e-02, 2.1484e-02, -6.7444e-03, ..., -6.0425e-03, + -1.3794e-02, -2.2095e-02], + [-1.9455e-03, -2.0508e-02, 5.4016e-03, ..., -4.1199e-03, + 8.1787e-03, 9.2163e-03], + [-1.6724e-02, 1.1292e-02, -1.0437e-02, ..., 1.8463e-03, + -1.7334e-02, -1.3123e-02], + ..., + [-3.1586e-03, 1.0315e-02, -2.2705e-02, ..., -3.3691e-02, + 9.5825e-03, 2.0885e-04], + [-1.2817e-02, 2.1362e-02, -1.9287e-02, ..., -3.4485e-03, + -1.2512e-02, 3.6163e-03], + [ 3.2959e-02, -1.4648e-02, -1.2695e-02, ..., -8.9111e-03, + 1.9226e-03, 2.3961e-05]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm.weight': tensor([ 9.6321e-05, -6.7234e-05, 7.1411e-03, 3.2501e-03, 2.7734e-01, + 7.8201e-05, 2.7061e-05, 1.3351e-04, -2.2793e-04, 4.1580e-04, + 1.1292e-03, 1.1780e-02, 1.6570e-05, 7.5989e-03, -4.6921e-04, + 2.0027e-05, 9.0599e-05, 2.0313e-04, -7.2479e-05, 2.8419e-04, + 3.2234e-04, 1.2207e-02, 8.9111e-03, 6.1035e-05, 6.2256e-03, + 5.8594e-03, 2.7148e-01, 3.1662e-04, 1.2085e-02, 3.6719e-01, + 7.7820e-04, 7.2002e-05, 1.3471e-05, -1.1015e-04, 2.2168e-01, + 1.1292e-03, 2.9182e-04, 2.5781e-01, 6.8054e-03, 7.1526e-05, + 1.1658e-02, 2.2602e-04, -1.0395e-04, 1.5926e-04, 6.2561e-04, + 5.7678e-03, -3.4714e-04, -1.6689e-04, -1.7834e-04, 9.0942e-03, + -1.8311e-04, 4.0894e-03, 4.3678e-04, 1.4687e-04, -1.9836e-04, + -3.7994e-03, 6.9141e-05, 2.9182e-04, 3.2425e-04, 6.3782e-03, + 7.4768e-04, 9.0599e-05, -7.2956e-05, 6.7711e-05, 4.4531e-01, + -1.2779e-04, 2.7466e-04, 2.8125e-01, 1.0559e-02, 3.6011e-03, + -2.2411e-04, -1.1206e-04, 1.0315e-02, 2.7771e-03, -7.5817e-05, + 3.0060e-03, -2.8133e-05, 9.4891e-05, -6.6757e-05, -1.1146e-05, + 8.9111e-03, -5.3406e-05, 2.8076e-03, 4.6692e-03, 3.5858e-03, + 4.1580e-04, 1.0376e-03, 2.5868e-05, 4.5395e-04, 1.2878e-02, + 9.8267e-03, 4.2188e-01, 3.0327e-04, 3.4523e-04, 3.9648e-01, + -3.7625e-07, 2.5757e-02, 9.4414e-05, 6.7902e-04, -4.9744e-03, + 6.4850e-05, -1.0395e-04, 3.8605e-03, -3.8338e-04, -1.8024e-04, + 1.5503e-02, -9.6798e-05, 1.0872e-04, 8.2397e-03, 2.0885e-04, + -8.2970e-05, -8.6784e-05, 1.5354e-04, 1.1475e-02, 2.4707e-01, + -1.0757e-03, -5.2214e-05, 7.1526e-05, 1.2875e-04, 1.3504e-03, + 5.4932e-04, 6.3896e-05, 1.5450e-04, 1.6724e-02, 7.0801e-03, + -6.6280e-05, 1.9264e-04, -1.5080e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_proj.weight': tensor([[-6.4850e-05, 2.6550e-03, 1.3809e-03, ..., -1.9531e-03, + -2.4719e-03, -3.0975e-03], + [ 8.3008e-03, 3.9368e-03, -5.0964e-03, ..., 2.9449e-03, + -3.4714e-04, -4.0245e-04], + [ 9.3384e-03, -3.9673e-03, -1.8387e-03, ..., 3.7689e-03, + 4.0894e-03, 5.5237e-03], + ..., + [-9.3384e-03, 7.9346e-04, -5.2185e-03, ..., 2.8992e-03, + 6.6376e-04, -2.8229e-03], + [-5.4550e-04, -4.8828e-04, -4.3945e-03, ..., -8.3542e-04, + 1.4771e-02, 4.4250e-03], + [-1.0620e-02, -2.4261e-03, 4.5471e-03, ..., 3.1586e-03, + 5.0049e-03, -7.6294e-03]], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm.weight': tensor([ 9.5844e-05, 2.9683e-05, 5.0049e-03, 3.2196e-03, 2.7539e-01, + -7.5817e-05, 6.0499e-06, -1.4210e-04, 2.2793e-04, 4.1580e-04, + 1.1673e-03, 1.2268e-02, -6.1512e-05, 1.0010e-02, 4.6921e-04, + -2.0981e-05, -7.4387e-05, 2.0409e-04, -7.0095e-05, -3.6163e-03, + 3.2997e-04, 1.2085e-02, 8.6670e-03, 7.8678e-05, 6.3477e-03, + 6.0120e-03, 2.7148e-01, 3.1471e-04, 1.1963e-02, 3.6914e-01, + 8.3160e-04, 6.8665e-05, 3.5763e-05, 1.1635e-04, 2.2168e-01, + -9.7275e-04, 2.9182e-04, 2.5586e-01, 6.8054e-03, 8.4877e-05, + 1.2146e-02, 2.2888e-04, -1.0395e-04, 1.5926e-04, 6.1417e-04, + 5.8594e-03, 3.4904e-04, 1.6689e-04, -1.6785e-04, 9.4604e-03, + -1.7834e-04, 3.3569e-04, 4.4250e-04, -1.4877e-04, -1.9741e-04, + -3.7842e-03, 6.5327e-05, 2.9373e-04, 3.2425e-04, 6.4087e-03, + 7.5150e-04, 8.9169e-05, -6.2943e-05, -6.8188e-05, 4.4531e-01, + -1.3065e-04, 2.7657e-04, 2.7930e-01, 1.0559e-02, 3.6316e-03, + -2.2316e-04, -1.0824e-04, 1.0254e-02, 2.7771e-03, -7.6771e-05, + 3.0365e-03, 4.0293e-05, 9.5844e-05, -5.7459e-05, -1.3292e-05, + 8.9111e-03, -4.3631e-05, -2.7008e-03, -1.4954e-02, 3.7994e-03, + -4.1389e-04, 7.8201e-04, 4.5776e-05, 4.5395e-04, 1.2878e-02, + 1.0193e-02, 4.1992e-01, 3.1281e-04, 3.4523e-04, 3.9453e-01, + -3.8147e-06, 2.5513e-02, 9.2983e-05, 6.6757e-04, -5.2490e-03, + 6.4373e-05, -9.8705e-05, 3.9978e-03, -3.8147e-04, 1.7929e-04, + 1.5198e-02, -9.7275e-05, 1.0777e-04, 7.7820e-03, 2.0981e-04, + 8.2493e-05, 8.6784e-05, 1.4400e-04, 1.1353e-02, 2.4902e-01, + -2.8839e-03, -5.1975e-05, -7.2479e-05, -1.2779e-04, 1.4114e-03, + 5.4932e-04, 6.4850e-05, 1.5450e-04, 1.6235e-02, 7.3547e-03, + -6.7234e-05, -1.9360e-04, 1.7047e-05], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.v_proj.weight': tensor([[-0.0231, -0.0107, -0.0337, ..., -0.0066, -0.0225, -0.0272], + [ 0.0203, 0.0035, 0.0232, ..., 0.0093, 0.0261, 0.0160], + [-0.0203, -0.0079, -0.0182, ..., -0.0113, -0.0298, -0.0155], + ..., + [-0.0154, -0.0115, -0.0270, ..., -0.0116, -0.0266, -0.0248], + [-0.0175, -0.0090, -0.0232, ..., -0.0089, -0.0234, -0.0113], + [-0.0018, 0.0097, 0.0084, ..., 0.0065, 0.0131, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.output_proj.weight': tensor([[-0.0008, -0.0026, 0.0054, ..., 0.0103, 0.0027, -0.0010], + [ 0.0120, 0.0098, 0.0007, ..., 0.0205, 0.0019, -0.0042], + [ 0.0159, -0.0123, 0.0170, ..., -0.0305, -0.0219, 0.0066], + ..., + [-0.0084, 0.0123, -0.0113, ..., 0.0315, 0.0244, 0.0019], + [-0.0166, 0.0074, -0.0073, ..., -0.0168, -0.0076, -0.0045], + [ 0.0211, -0.0052, 0.0154, ..., -0.0491, -0.0374, 0.0082]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img.weight': tensor([[-0.0664, 0.0305, 0.0151, ..., -0.0344, 0.0052, -0.0364], + [ 0.0220, -0.0026, -0.0115, ..., -0.0043, 0.0116, 0.0219], + [ 0.0317, -0.0347, 0.0117, ..., 0.0444, 0.0049, 0.0471], + ..., + [-0.0359, 0.0248, -0.0118, ..., -0.0067, 0.0192, -0.0393], + [-0.0102, -0.0244, 0.0275, ..., 0.0038, 0.0043, 0.0081], + [-0.0170, -0.0427, 0.0132, ..., 0.0203, -0.0505, 0.0225]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.v_img.weight': tensor([[-0.0050, -0.0378, -0.0359, ..., 0.0552, -0.0075, -0.0160], + [ 0.0047, 0.0142, 0.0134, ..., 0.0055, -0.0153, 0.0354], + [ 0.0094, 0.0040, 0.0057, ..., 0.0028, -0.0249, -0.0136], + ..., + [ 0.0068, -0.0125, 0.0049, ..., -0.0095, -0.0079, -0.0036], + [-0.0073, 0.0439, 0.0166, ..., 0.0248, 0.0134, -0.0187], + [ 0.0175, 0.0061, 0.0049, ..., 0.0410, -0.0096, -0.0030]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img.weight': tensor([[-0.0322, 0.0317, -0.0107, ..., -0.0425, 0.0376, 0.0168], + [-0.0216, 0.0142, 0.0038, ..., -0.0035, -0.0244, 0.0356], + [ 0.0071, 0.0165, -0.0200, ..., 0.0033, 0.0237, -0.0078], + ..., + [-0.0197, 0.0054, 0.0386, ..., 0.0181, 0.0109, 0.0211], + [ 0.0270, 0.0161, -0.0396, ..., 0.0226, -0.0047, -0.0374], + [-0.0085, -0.0095, 0.0172, ..., 0.0243, -0.0219, -0.0251]], + dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.20.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.20.mlp.layer1.weight': tensor([[ 0.0072, 0.0064, 0.0272, ..., -0.0002, 0.0192, -0.0106], + [ 0.0042, -0.0162, -0.0035, ..., 0.0043, -0.0120, 0.0083], + [-0.0075, 0.0099, -0.0003, ..., 0.0026, -0.0227, 0.0038], + ..., + [-0.0098, -0.0035, -0.0066, ..., -0.0109, 0.0040, 0.0101], + [-0.0040, -0.0004, 0.0376, ..., 0.0059, 0.0125, -0.0173], + [-0.0045, 0.0005, 0.0070, ..., 0.0098, 0.0086, -0.0035]], + dtype=torch.bfloat16), 'net.blocks.20.mlp.layer2.weight': tensor([[ 2.3804e-02, -8.9264e-04, 3.6316e-03, ..., -1.0254e-02, + -9.2163e-03, -2.8687e-03], + [-1.2939e-02, 1.7776e-03, -2.6123e-02, ..., 2.5635e-02, + 4.7913e-03, -3.0640e-02], + [ 1.0071e-02, -2.2217e-02, 2.0264e-02, ..., -5.2185e-03, + -1.8066e-02, -2.2461e-02], + ..., + [ 1.1169e-02, -1.3733e-02, 1.1587e-04, ..., -2.4414e-02, + -2.4048e-02, 1.1658e-02], + [ 1.3062e-02, 5.9509e-03, -4.6082e-03, ..., -4.0627e-04, + -7.2327e-03, -1.0681e-03], + [ 5.3711e-03, -4.4556e-03, -3.3112e-03, ..., 2.5034e-05, + -1.0254e-02, -2.8687e-03]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.1.weight': tensor([[-5.4688e-02, -5.3711e-02, -5.2734e-02, ..., 2.8419e-04, + 3.6240e-04, 3.3188e-04], + [ 5.6641e-02, 5.5908e-02, 5.5420e-02, ..., -1.6117e-04, + -1.6880e-04, -2.5940e-04], + [ 2.8198e-02, 2.6978e-02, 2.5513e-02, ..., 2.1458e-04, + 2.3270e-04, 2.4605e-04], + ..., + [-1.0059e-01, -9.9121e-02, -9.7168e-02, ..., -8.8692e-05, + -8.5831e-05, -1.4114e-04], + [-6.2012e-02, -6.0791e-02, -5.9326e-02, ..., 2.5177e-04, + 3.2234e-04, 3.6240e-04], + [-6.2988e-02, -6.1768e-02, -6.0791e-02, ..., 3.1662e-04, + 2.6894e-04, 2.9755e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_self_attn.2.weight': tensor([[ 0.0036, -0.0043, -0.0021, ..., 0.0071, 0.0036, 0.0041], + [ 0.0042, -0.0029, 0.0013, ..., 0.0032, 0.0043, 0.0042], + [ 0.0040, -0.0005, 0.0027, ..., -0.0015, 0.0026, 0.0027], + ..., + [ 0.0045, -0.0297, -0.0025, ..., -0.0088, 0.0236, 0.0210], + [ 0.0096, -0.0256, 0.0013, ..., -0.0045, 0.0212, 0.0204], + [-0.0240, 0.0339, -0.0046, ..., -0.0060, -0.0356, -0.0342]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.1.weight': tensor([[-7.0801e-02, -6.9824e-02, -6.8359e-02, ..., -1.4782e-04, + -1.9169e-04, -5.6267e-05], + [-6.3965e-02, -6.3477e-02, -6.2988e-02, ..., -1.4400e-04, + -1.7548e-04, -8.2016e-05], + [ 5.5420e-02, 5.3467e-02, 5.2002e-02, ..., 1.8787e-04, + 1.2016e-04, 5.4598e-05], + ..., + [-6.4453e-02, -6.3477e-02, -6.2256e-02, ..., -1.6975e-04, + -1.6499e-04, -2.0123e-04], + [ 6.7383e-02, 6.6406e-02, 6.4941e-02, ..., 8.6784e-05, + 8.4877e-05, 1.3447e-04], + [-1.5015e-02, -1.4343e-02, -1.3916e-02, ..., 1.6975e-04, + 2.7084e-04, 1.2398e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0168, 0.0110, -0.0130, ..., 0.0167, -0.0104, 0.0054], + [-0.0040, -0.0030, 0.0022, ..., -0.0031, 0.0037, -0.0054], + [-0.0058, 0.0054, -0.0055, ..., -0.0018, -0.0046, -0.0064], + ..., + [-0.0039, -0.0025, 0.0014, ..., -0.0022, 0.0024, 0.0161], + [-0.0156, -0.0009, 0.0019, ..., -0.0117, -0.0036, 0.0194], + [ 0.0009, 0.0205, -0.0187, ..., 0.0038, -0.0217, 0.0152]], + dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.1.weight': tensor([[-3.3203e-02, -3.1982e-02, -3.0640e-02, ..., 4.6730e-04, + 3.3379e-04, 3.9291e-04], + [ 1.0547e-01, 1.0205e-01, 1.0010e-01, ..., -3.9482e-04, + -4.9591e-04, -4.2534e-04], + [ 1.0352e-01, 1.0107e-01, 9.9121e-02, ..., -3.8338e-04, + -3.5858e-04, -3.5095e-04], + ..., + [-7.1289e-02, -6.8848e-02, -6.7871e-02, ..., 4.6921e-04, + 4.3106e-04, 4.8828e-04], + [-5.7129e-02, -5.6885e-02, -5.5664e-02, ..., -1.4496e-04, + -9.6321e-05, -4.6730e-05], + [-4.2236e-02, -4.0771e-02, -4.0283e-02, ..., 4.4441e-04, + 5.8365e-04, 5.3024e-04]], dtype=torch.bfloat16), 'net.blocks.20.adaln_modulation_mlp.2.weight': tensor([[-2.3193e-03, 4.1199e-03, 4.3945e-03, ..., -7.1106e-03, + -1.2283e-03, -1.1353e-02], + [-1.6937e-03, 1.0986e-03, 9.8419e-04, ..., -1.3275e-03, + 1.2875e-04, -3.6011e-03], + [-2.2125e-03, 2.1820e-03, 2.1362e-03, ..., -3.4790e-03, + -1.6093e-05, -5.1270e-03], + ..., + [-3.1738e-02, -6.2012e-02, -6.1523e-02, ..., 6.1035e-02, + 4.7852e-02, 4.8096e-02], + [-3.0151e-02, -6.9824e-02, -6.8848e-02, ..., 7.2266e-02, + 4.4189e-02, 5.9814e-02], + [ 4.9561e-02, 5.7861e-02, 5.7861e-02, ..., -5.7861e-02, + -5.8105e-02, -3.1494e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_proj.weight': tensor([[ 8.5449e-03, -7.8125e-03, 1.4221e-02, ..., -1.5869e-02, + -2.3346e-03, 1.9775e-02], + [ 5.9509e-04, -1.3530e-05, -1.0147e-03, ..., -1.7334e-02, + 2.0142e-03, -9.2773e-03], + [ 1.1658e-02, 7.2937e-03, -1.2329e-02, ..., -1.5259e-02, + 5.9509e-03, -8.4229e-03], + ..., + [ 2.9663e-02, 6.3477e-03, -1.2512e-02, ..., 1.4465e-02, + 4.4861e-03, 3.7079e-03], + [ 8.5449e-03, -4.2725e-03, -5.7373e-03, ..., -1.1475e-02, + -2.2095e-02, -6.7444e-03], + [ 1.4648e-02, -8.9111e-03, -1.0529e-03, ..., 1.0803e-02, + -6.5308e-03, 2.0264e-02]], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm.weight': tensor([1.6641, 1.0469, 0.6211, 1.3203, 0.2520, 1.2422, 1.4531, 1.1797, 1.1328, + 1.3125, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.0938, 1.3516, 1.3516, + 1.1172, 1.4141, 1.2344, 1.4375, 0.9805, 1.1875, 1.3281, 1.4766, 1.1016, + 1.2812, 1.0469, 1.2422, 1.4219, 1.2656, 1.2500, 1.2031, 1.3125, 1.2891, + 1.3672, 1.3516, 1.4062, 1.3906, 1.2031, 1.2578, 1.2578, 0.7500, 1.1953, + 1.7188, 1.6953, 1.2109, 1.4062, 1.2031, 1.1484, 1.2344, 1.2031, 1.2812, + 1.2578, 1.2812, 1.3750, 1.4062, 1.4531, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 0.5820, 1.3516, 0.3281, 0.8906, 0.2070, 1.1719, 1.0078, 1.4297, + 1.3828, 1.2891, 1.3516, 1.3438, 1.2734, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.8125, 1.7422, 1.6016, 1.6953, + 0.8516, 1.3047, 1.4609, 1.2109, 1.3438, 1.3359, 1.2422, 0.3672, 1.2656, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.9922, + 1.4297, 1.1172, 1.1562, 1.0234, 1.7266, 0.7109, 1.0156, 1.4297, 1.4297, + 1.3203, 1.4844, 0.4199, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.k_proj.weight': tensor([[-0.0220, -0.0149, 0.0129, ..., -0.0121, -0.0093, 0.0034], + [-0.0045, 0.0040, 0.0015, ..., 0.0026, -0.0064, -0.0023], + [-0.0251, 0.0006, 0.0066, ..., 0.0052, 0.0038, -0.0038], + ..., + [ 0.0234, 0.0126, -0.0020, ..., 0.0210, 0.0050, -0.0035], + [ 0.0123, 0.0140, -0.0062, ..., -0.0027, -0.0069, -0.0099], + [ 0.0074, -0.0155, 0.0065, ..., 0.0106, -0.0021, 0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm.weight': tensor([1.5312, 0.7734, 0.6289, 1.2266, 0.2500, 1.1797, 1.3359, 1.2734, 1.2891, + 1.3359, 1.2891, 1.3828, 1.2422, 1.2812, 1.2656, 1.1016, 1.3438, 1.3594, + 1.1172, 1.4141, 1.2344, 1.4375, 1.3438, 1.3438, 1.3047, 1.3594, 1.1172, + 1.2344, 1.1875, 1.2578, 1.4219, 1.2656, 1.2422, 1.2109, 1.3047, 1.2969, + 1.3672, 1.3516, 1.4062, 1.3828, 1.2031, 1.2578, 1.2578, 1.6562, 1.0312, + 1.5781, 1.6328, 1.1328, 1.2344, 1.1562, 1.1641, 1.2578, 1.2031, 1.2969, + 1.2891, 1.2812, 1.3672, 1.4062, 1.4453, 1.3359, 1.3359, 1.2969, 1.3828, + 0.9922, 1.0938, 1.3203, 0.2363, 1.0547, 0.2100, 1.1484, 1.0625, 1.3750, + 1.3984, 1.2656, 1.3516, 1.3359, 1.2812, 1.2578, 1.2812, 1.2969, 1.3047, + 1.2891, 1.3906, 1.2500, 1.3047, 1.1562, 1.7031, 1.6641, 1.1953, 1.6875, + 0.6875, 1.3516, 1.4141, 1.2031, 1.3359, 1.3359, 1.2422, 0.3691, 1.2578, + 1.3750, 1.2969, 1.2344, 1.3984, 1.4219, 1.3359, 1.3828, 1.3125, 1.4688, + 1.5625, 1.4141, 1.1641, 1.0469, 1.5781, 0.7500, 1.0234, 1.3828, 1.4297, + 1.3125, 1.4688, 0.4238, 1.0859, 0.8008, 1.2266, 1.2812, 1.3516, 1.3516, + 1.3047, 1.3125], dtype=torch.bfloat16), 'net.blocks.21.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.self_attn.v_proj.weight': tensor([[-0.0141, -0.0291, -0.0081, ..., -0.0070, 0.0182, 0.0161], + [-0.0034, 0.0154, -0.0049, ..., 0.0087, 0.0046, -0.0095], + [-0.0082, 0.0131, 0.0208, ..., 0.0079, -0.0151, -0.0024], + ..., + [-0.0106, 0.0048, -0.0024, ..., -0.0094, -0.0227, 0.0123], + [ 0.0075, 0.0176, 0.0099, ..., -0.0001, -0.0039, -0.0044], + [-0.0038, -0.0092, -0.0320, ..., 0.0162, -0.0153, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.21.self_attn.output_proj.weight': tensor([[-7.7209e-03, -1.2756e-02, 2.3651e-03, ..., 1.3000e-02, + 1.2451e-02, -9.2773e-03], + [ 2.2095e-02, 9.6436e-03, 5.4321e-03, ..., -9.6436e-03, + -2.0508e-02, 8.4839e-03], + [-1.3245e-02, 1.4954e-03, 6.8359e-03, ..., -9.6436e-03, + 1.6113e-02, -2.9785e-02], + ..., + [-4.6692e-03, 2.3926e-02, 1.1047e-02, ..., 9.4986e-04, + 1.1169e-02, -1.6357e-02], + [ 3.4637e-03, 2.7771e-03, -1.5259e-02, ..., -2.4048e-02, + 5.7220e-04, 1.0376e-03], + [ 1.4709e-02, -7.9632e-05, 3.4027e-03, ..., 1.5747e-02, + 2.0508e-02, 1.6479e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_proj.weight': tensor([[ 0.0062, -0.0032, -0.0080, ..., -0.0037, 0.0052, 0.0111], + [-0.0234, 0.0283, 0.0227, ..., 0.0071, 0.0142, 0.0134], + [ 0.0035, -0.0063, -0.0199, ..., -0.0016, -0.0095, 0.0124], + ..., + [ 0.0079, -0.0058, 0.0048, ..., -0.0187, -0.0075, -0.0084], + [-0.0018, 0.0074, -0.0347, ..., -0.0403, -0.0120, -0.0064], + [-0.0161, 0.0237, -0.0126, ..., -0.0261, -0.0105, 0.0144]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm.weight': tensor([ 1.2159e-04, 1.4782e-04, 2.7734e-01, 4.2188e-01, -2.8038e-04, + -2.1172e-04, 2.0142e-03, -1.2589e-04, 4.0817e-04, 9.7656e-03, + -4.2439e-05, 4.6692e-03, 5.7983e-04, 4.0770e-05, 1.0729e-04, + 2.3804e-03, -1.8024e-04, 4.8447e-04, 2.2827e-02, 6.5430e-02, + 4.2534e-04, 2.4033e-04, 1.2398e-04, 5.4121e-05, 1.8234e-03, + 2.2411e-04, 7.0095e-05, 2.7313e-03, 1.3542e-04, 8.3923e-05, + -5.3787e-04, -9.6321e-05, 3.7384e-03, -8.2016e-05, 6.5918e-03, + 1.8555e-02, -1.0109e-04, 3.3379e-05, 1.5030e-03, -5.0783e-05, + 2.9755e-03, 2.8014e-05, 1.2398e-04, 1.9360e-04, 4.1485e-05, + -2.1267e-04, 2.3041e-03, 1.4114e-04, -8.9645e-05, 6.9885e-03, + 4.9609e-01, 4.8340e-02, 7.0190e-03, -9.6321e-05, 3.9368e-03, + -2.9182e-04, 4.4250e-04, 4.2236e-02, -4.7922e-05, 6.4087e-03, + 1.7762e-05, -9.3937e-05, -2.1839e-04, -1.4687e-04, 1.2970e-04, + 2.1553e-04, 8.1539e-05, 9.2773e-02, 1.9836e-03, 2.1172e-04, + -1.5354e-04, 7.7248e-05, 3.4943e-03, 1.5381e-02, 2.0508e-02, + 9.1553e-03, -1.3447e-04, -1.1921e-04, -1.4877e-04, 1.0071e-02, + 4.7461e-01, 1.3245e-02, 8.7738e-04, 2.9297e-01, 1.0443e-04, + 3.4570e-01, 2.6093e-03, 3.4332e-05, 2.4567e-03, 3.9453e-01, + 5.9891e-04, 2.2430e-03, -2.9945e-04, -8.2970e-05, 2.1820e-03, + -3.3569e-04, 2.3365e-04, 4.6387e-03, -5.7697e-05, 7.1106e-03, + 8.4877e-05, -1.3885e-03, -2.8372e-05, -1.1158e-04, 1.0742e-02, + 1.0620e-02, 1.4114e-04, 1.1414e-02, 1.5545e-04, 5.9843e-05, + 4.1504e-03, -1.2970e-04, 1.9409e-02, 1.7929e-04, 1.1673e-03, + 1.4355e-01, 1.9550e-04, -9.5844e-05, 1.4496e-04, 8.3923e-05, + 2.7466e-03, 2.9297e-01, 4.9316e-02, 8.8501e-03, 2.9373e-04, + 4.0039e-02, 9.3842e-04, -1.4305e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_proj.weight': tensor([[ 1.1520e-03, 3.7537e-03, 7.4768e-03, ..., -8.9645e-04, + 1.2741e-03, -4.9744e-03], + [-1.7090e-02, 2.8839e-03, 1.3885e-03, ..., 3.0327e-04, + 6.7139e-03, -8.3008e-03], + [-3.5477e-04, -6.3477e-03, -6.4697e-03, ..., -7.2021e-03, + 1.0529e-03, 9.6512e-04], + ..., + [-7.7820e-03, -2.0123e-04, -6.0730e-03, ..., 1.8311e-04, + 2.6093e-03, -7.3242e-03], + [-1.8463e-03, -8.6212e-04, 1.6235e-02, ..., 9.5215e-03, + -6.1035e-03, 5.9204e-03], + [ 3.6621e-03, 5.1856e-06, -2.3499e-03, ..., 6.7749e-03, + -7.1716e-03, -3.2234e-04]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm.weight': tensor([ 1.2159e-04, 1.4687e-04, 2.7734e-01, 4.1602e-01, -2.8229e-04, + 2.1648e-04, 2.0294e-03, 1.2589e-04, 3.4904e-04, 9.8267e-03, + 4.4346e-05, 4.6082e-03, 5.6839e-04, 2.2888e-05, -1.0681e-04, + 2.5787e-03, -1.8024e-04, -4.8828e-04, 2.2583e-02, -1.8616e-03, + 4.2343e-04, 2.3460e-04, 1.2589e-04, 5.6505e-05, -1.2817e-03, + 2.2411e-04, 6.0081e-05, 2.5024e-03, 1.5354e-04, 8.6308e-05, + -5.4169e-04, -9.5844e-05, 3.6926e-03, -8.6784e-05, 6.6223e-03, + -4.3945e-03, -1.0061e-04, 2.0695e-04, 1.4954e-03, -5.0545e-05, + 2.2888e-03, 2.8253e-05, 1.8311e-04, 1.9264e-04, 4.2915e-05, + -2.1267e-04, 2.3499e-03, 1.4782e-04, -8.9169e-05, 6.9885e-03, + 4.9414e-01, -1.9684e-03, 7.1411e-03, 9.5367e-05, 3.9673e-03, + 2.9182e-04, 1.7929e-04, 4.2725e-02, -5.4121e-05, 6.3782e-03, + -1.2934e-05, -9.5367e-05, 2.2030e-04, 1.2970e-04, 1.2875e-04, + 1.6594e-04, 7.9632e-05, 1.3199e-03, 1.9455e-03, -2.0981e-04, + -1.5163e-04, 7.9632e-05, 6.3324e-04, 1.5381e-02, 2.0142e-02, + 9.0332e-03, -1.3542e-04, -1.2016e-04, 1.4782e-04, 1.0071e-02, + 4.7461e-01, 1.2695e-02, 7.7438e-04, 2.9102e-01, -1.0633e-04, + 3.4570e-01, 2.7313e-03, -3.3855e-05, 1.8234e-03, 3.9844e-01, + 6.7520e-04, 2.1820e-03, -2.9945e-04, -8.3923e-05, 2.1667e-03, + 3.0136e-04, 2.5940e-04, 4.6692e-03, -5.7936e-05, -3.3722e-03, + -8.2970e-05, 1.3580e-03, -3.0041e-05, -1.1158e-04, -8.7280e-03, + 1.1292e-02, -1.3924e-04, 1.0925e-02, 1.5545e-04, 7.2956e-05, + 4.1504e-03, 1.6499e-04, 1.9409e-02, 1.8215e-04, 1.1139e-03, + -2.2125e-03, 1.9550e-04, 9.8228e-05, 1.4496e-04, 8.2970e-05, + -6.1035e-04, 2.9297e-01, 4.8828e-02, 8.8501e-03, 2.9373e-04, + 4.0283e-02, 8.6975e-04, -1.3828e-04], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.v_proj.weight': tensor([[ 0.0096, 0.0104, 0.0167, ..., 0.0083, 0.0183, 0.0140], + [ 0.0130, 0.0102, 0.0159, ..., 0.0081, 0.0095, 0.0015], + [ 0.0157, 0.0106, 0.0206, ..., 0.0089, 0.0155, -0.0034], + ..., + [-0.0033, 0.0137, 0.0018, ..., 0.0055, 0.0126, 0.0243], + [-0.0021, 0.0156, 0.0194, ..., 0.0082, 0.0085, 0.0094], + [-0.0033, -0.0115, -0.0056, ..., -0.0103, -0.0269, -0.0177]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.output_proj.weight': tensor([[-0.0006, -0.0076, -0.0009, ..., -0.0103, 0.0084, 0.0073], + [ 0.0081, -0.0019, -0.0016, ..., -0.0024, 0.0019, 0.0062], + [ 0.0019, 0.0019, -0.0046, ..., -0.0189, -0.0200, 0.0184], + ..., + [ 0.0155, -0.0056, -0.0019, ..., 0.0105, 0.0067, -0.0130], + [ 0.0099, -0.0121, -0.0125, ..., -0.0014, 0.0014, -0.0041], + [-0.0190, 0.0040, -0.0040, ..., -0.0181, 0.0046, 0.0173]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img.weight': tensor([[-3.4668e-02, 6.2866e-03, 1.9287e-02, ..., -2.2095e-02, + -1.4771e-02, -2.1118e-02], + [ 1.3062e-02, 1.1658e-02, 2.5269e-02, ..., 2.0386e-02, + -2.3193e-02, -3.2227e-02], + [ 2.5513e-02, -4.8523e-03, 4.1809e-03, ..., -5.7068e-03, + 5.4016e-03, 3.3691e-02], + ..., + [ 4.2236e-02, 2.4414e-03, 3.5156e-02, ..., -3.0396e-02, + -4.7207e-05, -2.0386e-02], + [ 3.0151e-02, 1.5503e-02, 9.9487e-03, ..., -7.2327e-03, + -2.5391e-02, -2.4536e-02], + [-6.4087e-04, -6.2561e-03, -2.0599e-03, ..., 4.1504e-03, + -1.9653e-02, -2.8198e-02]], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.v_img.weight': tensor([[-0.0200, -0.0042, 0.0266, ..., 0.0155, 0.0187, 0.0047], + [-0.0118, -0.0148, 0.0092, ..., 0.0408, -0.0205, -0.0097], + [-0.0189, 0.0205, 0.0036, ..., -0.0172, -0.0092, 0.0231], + ..., + [-0.0295, -0.0136, -0.0036, ..., 0.0042, -0.0048, 0.0101], + [-0.0048, -0.0140, 0.0317, ..., -0.0152, -0.0459, -0.0078], + [-0.0015, -0.0043, -0.0259, ..., -0.0030, -0.0141, -0.0115]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img.weight': tensor([[ 0.0040, 0.0064, -0.0140, ..., 0.0242, -0.0040, -0.0386], + [-0.0142, 0.0068, -0.0101, ..., -0.0369, -0.0151, 0.0157], + [-0.0267, -0.0045, -0.0074, ..., -0.0126, 0.0182, -0.0115], + ..., + [ 0.0172, -0.0038, 0.0144, ..., 0.0115, -0.0024, 0.0530], + [ 0.0239, 0.0265, -0.0095, ..., -0.0023, -0.0033, -0.0137], + [-0.0219, -0.0396, 0.0039, ..., -0.0054, -0.0037, -0.0132]], + dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.21.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.21.mlp.layer1.weight': tensor([[ 0.0120, -0.0020, 0.0098, ..., -0.0159, 0.0151, -0.0081], + [-0.0369, 0.0084, 0.0097, ..., 0.0124, -0.0008, 0.0068], + [ 0.0369, -0.0084, -0.0209, ..., 0.0132, 0.0047, 0.0151], + ..., + [-0.0105, -0.0113, -0.0176, ..., 0.0198, -0.0055, -0.0192], + [-0.0157, -0.0101, -0.0175, ..., -0.0020, -0.0021, 0.0183], + [ 0.0149, -0.0186, 0.0030, ..., 0.0131, 0.0078, -0.0067]], + dtype=torch.bfloat16), 'net.blocks.21.mlp.layer2.weight': tensor([[ 0.0091, -0.0061, 0.0248, ..., -0.0049, 0.0332, -0.0056], + [-0.0129, 0.0028, 0.0166, ..., -0.0013, -0.0344, 0.0065], + [ 0.0036, -0.0035, -0.0200, ..., -0.0063, -0.0070, -0.0003], + ..., + [-0.0052, -0.0068, 0.0087, ..., 0.0021, 0.0115, 0.0068], + [ 0.0146, -0.0271, -0.0104, ..., -0.0094, 0.0101, 0.0112], + [ 0.0083, -0.0027, -0.0091, ..., -0.0039, -0.0039, 0.0150]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.1.weight': tensor([[ 0.0747, 0.0737, 0.0728, ..., -0.0004, -0.0005, -0.0004], + [-0.0688, -0.0674, -0.0659, ..., 0.0009, 0.0008, 0.0008], + [-0.0679, -0.0659, -0.0645, ..., 0.0009, 0.0009, 0.0008], + ..., + [ 0.0649, 0.0625, 0.0615, ..., -0.0009, -0.0009, -0.0009], + [-0.0461, -0.0459, -0.0461, ..., 0.0002, 0.0002, 0.0002], + [ 0.0649, 0.0635, 0.0620, ..., -0.0008, -0.0009, -0.0009]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_self_attn.2.weight': tensor([[-3.7537e-03, 2.5024e-03, 2.7618e-03, ..., -1.9073e-03, + 3.8757e-03, -1.6861e-03], + [ 1.7624e-03, -1.5259e-03, -1.5488e-03, ..., 1.5259e-03, + -4.0245e-04, 1.6174e-03], + [-7.6771e-05, -3.9062e-03, -3.9062e-03, ..., 3.9062e-03, + 1.2970e-03, 3.9978e-03], + ..., + [-3.9307e-02, 4.9561e-02, 4.9805e-02, ..., -4.8584e-02, + 2.9297e-02, -4.3701e-02], + [-3.0396e-02, 1.9287e-02, 1.9653e-02, ..., -1.5503e-02, + 2.9785e-02, -1.0193e-02], + [-4.1504e-02, 2.3804e-02, 2.3682e-02, ..., -1.9531e-02, + 4.2236e-02, -1.2695e-02]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.1.weight': tensor([[ 2.2339e-02, 2.2339e-02, 2.2583e-02, ..., 2.3651e-04, + 1.5450e-04, 2.5177e-04], + [ 2.3926e-02, 2.3804e-02, 2.3682e-02, ..., 2.4033e-04, + 2.2507e-04, 2.2602e-04], + [-7.2266e-02, -6.9824e-02, -6.8848e-02, ..., -1.7166e-04, + -1.5354e-04, -1.6117e-04], + ..., + [-4.9805e-02, -4.9072e-02, -4.7607e-02, ..., 9.3460e-05, + 1.7071e-04, 1.3828e-04], + [-6.7383e-02, -6.5918e-02, -6.3965e-02, ..., -7.4387e-05, + -2.0218e-04, -1.3828e-04], + [ 4.7119e-02, 4.6387e-02, 4.5410e-02, ..., -2.5177e-04, + -1.6499e-04, -2.2411e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_cross_attn.2.weight': tensor([[-0.0043, 0.0009, -0.0095, ..., 0.0006, -0.0099, -0.0006], + [-0.0010, -0.0013, -0.0053, ..., 0.0010, -0.0047, -0.0010], + [-0.0009, -0.0010, -0.0020, ..., -0.0024, -0.0006, 0.0024], + ..., + [ 0.0371, 0.0371, -0.0342, ..., 0.0216, -0.0342, -0.0219], + [ 0.0292, 0.0303, -0.0272, ..., 0.0182, -0.0250, -0.0184], + [-0.0233, -0.0245, 0.0186, ..., -0.0103, 0.0178, 0.0107]], + dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.1.weight': tensor([[ 1.0791e-01, 1.0645e-01, 1.0400e-01, ..., -4.3678e-04, + -4.5395e-04, -4.4632e-04], + [-5.3223e-02, -5.2002e-02, -5.0537e-02, ..., 4.1389e-04, + 4.5204e-04, 4.9591e-04], + [-9.5703e-02, -9.3750e-02, -9.0820e-02, ..., 4.1580e-04, + 4.7112e-04, 4.9210e-04], + ..., + [ 6.9824e-02, 6.7871e-02, 6.7383e-02, ..., -6.3419e-05, + -1.7452e-04, -1.0014e-04], + [-3.8086e-02, -3.7842e-02, -3.7354e-02, ..., -2.1744e-04, + -6.6280e-05, -1.0443e-04], + [-1.2268e-02, -1.1780e-02, -1.1047e-02, ..., 1.7738e-04, + 2.6512e-04, 2.8610e-04]], dtype=torch.bfloat16), 'net.blocks.21.adaln_modulation_mlp.2.weight': tensor([[ 2.9945e-04, 7.9727e-04, -7.5531e-04, ..., -1.3199e-03, + 5.4550e-04, -4.2725e-04], + [ 1.0376e-03, -1.4877e-03, -1.2665e-03, ..., 4.4250e-04, + -1.1539e-04, -4.9591e-04], + [ 1.7319e-03, -6.6757e-05, -2.3346e-03, ..., -3.4142e-04, + 4.2152e-04, -7.9727e-04], + ..., + [-7.5684e-02, 3.1738e-02, 1.0303e-01, ..., -3.9795e-02, + 3.4424e-02, -1.8188e-02], + [-5.6641e-02, -1.3123e-02, 6.7383e-02, ..., -4.3213e-02, + 4.1748e-02, -3.5645e-02], + [ 6.6895e-02, 1.4282e-02, -8.1055e-02, ..., 5.2979e-02, + -5.2002e-02, 4.3213e-02]], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_proj.weight': tensor([[-0.0242, -0.0101, 0.0137, ..., -0.0110, -0.0031, 0.0038], + [ 0.0256, -0.0049, -0.0115, ..., 0.0237, 0.0145, -0.0082], + [-0.0295, 0.0051, 0.0231, ..., -0.0033, -0.0012, 0.0065], + ..., + [ 0.0073, 0.0190, 0.0108, ..., 0.0032, -0.0015, 0.0033], + [-0.0089, -0.0085, 0.0239, ..., 0.0121, 0.0028, -0.0014], + [ 0.0038, 0.0045, -0.0111, ..., -0.0004, -0.0038, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm.weight': tensor([1.2422, 2.0312, 0.8750, 0.7812, 0.9844, 0.7383, 0.8086, 1.0469, 1.1953, + 0.2412, 1.2891, 0.9688, 1.2656, 1.2812, 1.2031, 1.2578, 1.2422, 1.3906, + 0.1963, 1.4141, 1.0547, 1.3203, 0.7344, 0.8555, 1.4766, 1.0625, 0.9961, + 1.5781, 1.1562, 1.3906, 1.2812, 1.3594, 1.2812, 1.3594, 1.3750, 1.1562, + 1.1797, 1.3203, 1.2031, 1.3828, 1.3047, 1.3906, 1.3203, 0.9883, 1.2266, + 1.1094, 1.2344, 0.7383, 1.7109, 1.3672, 1.1797, 1.1406, 1.3594, 1.2891, + 1.3281, 1.3047, 1.4062, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3672, 1.0234, 0.5664, 1.1094, 1.1641, 1.3750, 0.6562, 0.4883, 1.2344, + 1.3516, 1.4609, 1.3281, 1.3359, 1.2891, 1.2188, 1.2969, 1.2422, 1.2969, + 1.3125, 1.2734, 1.2500, 0.8750, 0.8906, 1.7812, 1.1641, 1.8438, 1.2422, + 0.7422, 1.3906, 1.5391, 1.2812, 1.3750, 1.2656, 1.3359, 1.2812, 1.0078, + 1.4141, 1.3516, 1.3594, 1.2578, 1.4062, 1.1953, 1.3750, 1.3125, 1.6406, + 1.2656, 1.2031, 1.7656, 0.8672, 1.3438, 1.1484, 0.9375, 1.4766, 1.3203, + 1.3672, 1.3203, 1.3203, 1.3359, 1.3828, 1.3516, 1.0703, 1.1953, 1.3828, + 1.2266, 1.2891], dtype=torch.bfloat16), 'net.blocks.22.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.k_proj.weight': tensor([[-0.0156, 0.0066, 0.0320, ..., 0.0154, 0.0177, 0.0065], + [ 0.0164, -0.0037, -0.0090, ..., 0.0070, 0.0166, -0.0044], + [-0.0025, 0.0017, -0.0048, ..., 0.0106, 0.0040, -0.0068], + ..., + [-0.0015, 0.0040, 0.0186, ..., 0.0046, 0.0149, -0.0004], + [-0.0027, -0.0020, -0.0107, ..., -0.0010, -0.0053, 0.0015], + [-0.0027, -0.0114, -0.0139, ..., 0.0020, -0.0118, -0.0136]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm.weight': tensor([1.6250, 1.4375, 0.9219, 0.6719, 0.9961, 0.7500, 0.8828, 1.0547, 1.2812, + 0.2559, 1.3359, 0.9688, 1.2734, 1.2812, 1.2109, 1.2578, 1.2422, 1.3516, + 0.1973, 1.4141, 1.0469, 1.3203, 1.1797, 1.1172, 1.3984, 1.1016, 0.9648, + 1.4922, 1.2188, 1.3594, 1.3281, 1.3594, 1.2812, 1.2969, 1.3750, 1.1562, + 1.1797, 1.3281, 1.2109, 1.3906, 1.3047, 1.3906, 1.3203, 1.2891, 0.7227, + 0.7969, 1.2891, 0.9023, 1.4844, 1.3828, 1.0938, 1.2500, 1.3438, 1.2969, + 1.3359, 1.3047, 1.3984, 0.9453, 1.2812, 1.2656, 1.3516, 1.3906, 1.3438, + 1.3750, 0.9453, 1.0703, 1.0781, 1.1172, 1.3438, 0.5273, 0.3223, 1.2266, + 1.3359, 1.4609, 1.3203, 1.3516, 1.2812, 1.2188, 1.2969, 1.2422, 1.2969, + 1.2734, 1.2656, 1.2578, 0.8750, 0.8906, 1.5312, 0.6016, 1.6875, 1.0391, + 0.7852, 1.5078, 1.4766, 1.3672, 1.3516, 1.2578, 1.3438, 1.2422, 1.0156, + 1.4141, 1.3516, 1.3594, 1.2500, 1.4141, 1.1953, 1.3750, 1.3125, 1.5469, + 1.6250, 1.3359, 1.5859, 0.6719, 1.2578, 1.0703, 0.9727, 1.5234, 1.3203, + 1.3516, 1.3203, 1.3125, 1.3359, 1.3828, 1.3594, 1.0781, 1.2031, 1.3828, + 1.2344, 1.2812], dtype=torch.bfloat16), 'net.blocks.22.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.self_attn.v_proj.weight': tensor([[ 0.0145, -0.0042, 0.0036, ..., -0.0038, -0.0134, -0.0038], + [-0.0093, -0.0036, -0.0050, ..., 0.0128, -0.0057, -0.0067], + [-0.0084, -0.0114, -0.0058, ..., -0.0106, -0.0034, -0.0020], + ..., + [-0.0006, 0.0027, -0.0186, ..., -0.0043, -0.0111, -0.0023], + [-0.0049, 0.0271, 0.0231, ..., 0.0249, 0.0039, 0.0037], + [-0.0079, 0.0157, 0.0019, ..., -0.0140, -0.0108, -0.0013]], + dtype=torch.bfloat16), 'net.blocks.22.self_attn.output_proj.weight': tensor([[ 0.0294, -0.0034, 0.0058, ..., 0.0057, -0.0110, -0.0069], + [-0.0142, 0.0074, -0.0098, ..., 0.0119, -0.0101, -0.0410], + [ 0.0058, -0.0122, -0.0003, ..., -0.0306, -0.0002, 0.0041], + ..., + [-0.0154, 0.0237, -0.0008, ..., -0.0067, 0.0001, 0.0004], + [-0.0210, -0.0153, -0.0093, ..., -0.0262, 0.0009, 0.0148], + [ 0.0199, 0.0117, -0.0021, ..., 0.0033, 0.0034, 0.0024]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_proj.weight': tensor([[ 0.0079, -0.0049, 0.0135, ..., -0.0004, -0.0042, 0.0015], + [ 0.0027, 0.0037, 0.0413, ..., -0.0332, 0.0228, -0.0026], + [-0.0137, -0.0013, -0.0161, ..., 0.0048, 0.0223, 0.0186], + ..., + [-0.0035, 0.0057, -0.0041, ..., 0.0038, -0.0075, -0.0070], + [ 0.0297, -0.0417, 0.0007, ..., 0.0200, 0.0037, -0.0215], + [-0.0160, 0.0112, 0.0036, ..., -0.0056, -0.0065, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm.weight': tensor([ 4.7684e-05, -5.0354e-04, 2.6398e-03, -5.3787e-04, 1.6403e-04, + -1.5640e-04, 3.2812e-01, -4.0293e-05, 3.1891e-03, 9.3460e-05, + -7.0095e-05, -6.7234e-05, 8.8215e-05, 4.9133e-03, 8.1177e-03, + -1.0254e-02, -3.8910e-03, 7.4768e-04, -2.8419e-04, 1.0300e-04, + -2.1973e-03, -7.1526e-05, 6.1951e-03, -5.3787e-04, 2.1973e-03, + -3.4094e-05, 1.2451e-02, 2.0862e-05, -4.2343e-04, 1.8387e-03, + -3.0041e-05, -4.5538e-05, -3.0365e-03, -8.4839e-03, 5.0659e-03, + -5.3406e-05, 5.3516e-01, 3.3081e-06, 7.1526e-05, -4.5598e-06, + 5.6763e-03, -2.0599e-04, -9.3579e-06, 8.1539e-05, 4.6007e-07, + 2.2168e-01, 7.4005e-04, -1.5717e-03, -2.3246e-05, -2.5034e-05, + -5.2214e-05, 9.0599e-05, 7.2266e-02, -8.0585e-05, 2.8014e-05, + -2.5940e-03, -2.1935e-05, -1.2815e-05, 1.7929e-04, -5.6267e-05, + -6.1691e-06, -3.9368e-03, -1.9646e-04, 4.9438e-03, -2.5392e-05, + 1.7548e-03, 1.3550e-02, 4.4346e-05, 2.9802e-05, -1.6117e-04, + 3.8862e-05, 1.0834e-03, -1.0834e-03, 5.2643e-04, 9.4223e-04, + 8.9407e-06, -7.9727e-04, -4.6730e-05, 1.6689e-04, 5.1575e-03, + 8.9264e-04, 1.7456e-02, 4.1962e-04, 1.0061e-04, 2.7299e-05, + -1.5378e-05, 3.3379e-04, -1.4067e-05, -5.3406e-05, -3.0160e-05, + 3.3826e-06, -1.1015e-04, 5.2261e-04, 1.7822e-02, -7.3761e-07, + -1.7166e-04, -2.0742e-05, -5.4169e-04, -1.5545e-04, 3.0151e-02, + 5.6458e-04, 2.3937e-04, 6.6280e-05, 2.3603e-05, 2.7061e-05, + -2.7084e-04, 3.0078e-01, 2.4121e-01, 2.4676e-05, -6.5613e-04, + -5.3942e-06, 3.7766e-04, 2.8849e-05, 2.8419e-04, 1.2939e-02, + -9.2506e-05, 1.3428e-03, 5.7459e-05, 1.3161e-04, 1.8024e-04, + 3.2187e-05, -7.5623e-07, -1.5497e-05, -1.4114e-03, 9.3460e-05, + 1.8768e-03, -1.7643e-04, 5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_proj.weight': tensor([[-0.0056, 0.0004, -0.0074, ..., 0.0005, -0.0042, -0.0012], + [ 0.0113, -0.0044, -0.0020, ..., 0.0029, 0.0082, -0.0020], + [-0.0208, 0.0008, -0.0070, ..., 0.0032, -0.0077, 0.0007], + ..., + [ 0.0140, -0.0034, 0.0002, ..., -0.0043, -0.0135, -0.0046], + [-0.0189, -0.0023, -0.0026, ..., -0.0001, 0.0025, -0.0003], + [ 0.0145, 0.0127, -0.0087, ..., -0.0076, 0.0019, -0.0016]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm.weight': tensor([ 4.1962e-05, -4.9591e-04, 2.6398e-03, 5.4169e-04, -1.7166e-04, + 1.5545e-04, 3.2422e-01, 4.0293e-05, -3.1738e-03, 9.4891e-05, + 7.0572e-05, -7.7248e-05, 9.7275e-05, -2.6001e-02, 8.1787e-03, + -1.0315e-02, -3.8757e-03, 7.5531e-04, 8.1062e-05, -1.0300e-04, + -2.1667e-03, 7.1526e-05, 6.1646e-03, 5.3406e-04, -2.1973e-03, + -3.2037e-06, 1.2207e-02, 7.3433e-05, -4.2534e-04, -3.8086e-02, + -2.6822e-05, -5.8174e-05, 3.0365e-03, 8.4839e-03, -5.1575e-03, + 5.3644e-05, 5.3125e-01, -3.3081e-06, 7.2002e-05, 1.2144e-06, + 5.6458e-03, 2.0218e-04, 1.2636e-05, 5.0306e-05, 7.9721e-07, + 2.1973e-01, 7.2861e-04, 1.5564e-03, 2.3842e-05, 3.2663e-05, + -5.6982e-05, -9.2506e-05, 7.2266e-02, 7.9632e-05, 2.6822e-05, + -2.5940e-03, 2.2531e-05, 6.9737e-06, 1.6403e-04, 4.2200e-05, + -6.1095e-06, -2.8809e-02, -1.9550e-04, 4.8828e-03, -2.8133e-05, + 1.7929e-03, 1.3550e-02, -4.5300e-05, 2.7061e-05, -1.5640e-04, + 3.8385e-05, -1.0834e-03, 1.0834e-03, 5.2643e-04, 9.1934e-04, + 7.8678e-06, 7.8964e-04, -2.6345e-05, -2.5940e-04, -5.1575e-03, + -8.8882e-04, 1.7212e-02, -5.2261e-04, 1.0252e-04, 3.6322e-07, + 1.5497e-05, 3.3569e-04, -4.1910e-08, -1.7357e-04, -2.2292e-05, + -1.1921e-05, -1.1873e-04, -5.3406e-04, 1.7334e-02, -1.2890e-06, + 1.6308e-04, 1.5616e-05, -5.3787e-04, 5.9128e-05, 3.0151e-02, + 5.6458e-04, 2.4796e-04, 7.0572e-05, -2.3127e-05, -2.6107e-05, + 2.6703e-04, 3.0078e-01, 2.4023e-01, -2.5988e-05, -1.7090e-02, + -4.4107e-06, -4.1199e-04, -2.1100e-05, -2.7275e-04, 1.2878e-02, + 9.2983e-05, 1.2054e-03, -5.7220e-05, 1.3161e-04, 1.8406e-04, + -3.2187e-05, -1.5018e-08, 1.5020e-05, -1.4267e-03, 9.3460e-05, + -1.8555e-02, -1.7548e-04, -5.4932e-03], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.v_proj.weight': tensor([[-0.0031, -0.0040, -0.0183, ..., -0.0035, -0.0126, -0.0048], + [ 0.0192, 0.0082, 0.0381, ..., 0.0087, 0.0164, 0.0214], + [ 0.0031, 0.0039, 0.0195, ..., 0.0057, 0.0131, -0.0002], + ..., + [ 0.0186, 0.0046, 0.0410, ..., 0.0062, 0.0229, 0.0151], + [ 0.0156, 0.0057, 0.0425, ..., 0.0064, 0.0266, 0.0219], + [ 0.0173, 0.0056, 0.0437, ..., 0.0069, 0.0253, 0.0233]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.output_proj.weight': tensor([[ 0.0069, -0.0033, 0.0118, ..., -0.0095, -0.0144, -0.0124], + [-0.0040, -0.0052, -0.0096, ..., 0.0045, -0.0011, 0.0064], + [-0.0069, 0.0003, -0.0024, ..., 0.0173, 0.0188, 0.0176], + ..., + [ 0.0172, -0.0265, -0.0074, ..., 0.0027, 0.0032, 0.0064], + [ 0.0026, -0.0059, -0.0044, ..., -0.0014, -0.0008, -0.0014], + [ 0.0066, 0.0132, 0.0153, ..., -0.0065, -0.0056, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img.weight': tensor([[-3.3417e-03, 1.6846e-02, 1.9409e-02, ..., 3.4912e-02, + -4.1992e-02, 2.0142e-03], + [ 4.3945e-03, 2.7344e-02, 1.4526e-02, ..., -1.8433e-02, + -3.5248e-03, 1.5442e-02], + [ 1.4465e-02, -3.7231e-03, -7.3242e-03, ..., 2.4536e-02, + 3.5889e-02, 5.8289e-03], + ..., + [ 4.3213e-02, 1.4526e-02, -3.1982e-02, ..., 8.8215e-05, + 4.2114e-03, 2.6611e-02], + [ 2.2949e-02, 1.2573e-02, -1.9897e-02, ..., -7.5989e-03, + -3.1250e-02, -4.5654e-02], + [ 4.9133e-03, 1.0910e-03, -6.8359e-03, ..., 4.9133e-03, + -1.0437e-02, -1.5503e-02]], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.v_img.weight': tensor([[-0.0209, 0.0160, 0.0248, ..., 0.0115, 0.0388, 0.0359], + [-0.0042, 0.0121, -0.0019, ..., 0.0123, -0.0046, 0.0068], + [-0.0277, 0.0011, -0.0181, ..., -0.0003, -0.0036, 0.0330], + ..., + [-0.0234, -0.0278, -0.0096, ..., 0.0035, 0.0513, -0.0109], + [ 0.0211, 0.0166, 0.0171, ..., -0.0376, -0.0186, -0.0110], + [-0.0282, -0.0046, 0.0119, ..., -0.0055, -0.0038, 0.0243]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img.weight': tensor([[ 0.0272, -0.0033, -0.0128, ..., 0.0140, 0.0308, 0.0059], + [ 0.0120, -0.0332, -0.0320, ..., -0.0075, -0.0140, -0.0032], + [ 0.0312, 0.0488, -0.0080, ..., -0.0189, -0.0142, 0.0188], + ..., + [-0.0294, -0.0400, 0.0170, ..., -0.0089, 0.0295, 0.0505], + [-0.0159, 0.0029, -0.0311, ..., 0.0124, 0.0059, 0.0081], + [ 0.0040, 0.0198, 0.0141, ..., -0.0222, -0.0143, 0.0164]], + dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.22.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.22.mlp.layer1.weight': tensor([[ 0.0073, 0.0028, -0.0104, ..., 0.0238, -0.0141, -0.0018], + [ 0.0057, 0.0003, 0.0007, ..., -0.0136, 0.0119, 0.0038], + [ 0.0017, -0.0120, -0.0126, ..., -0.0098, 0.0038, -0.0201], + ..., + [ 0.0264, -0.0137, 0.0228, ..., -0.0099, 0.0096, -0.0038], + [ 0.0079, -0.0037, -0.0067, ..., 0.0109, 0.0123, -0.0181], + [ 0.0150, -0.0376, 0.0038, ..., 0.0013, -0.0036, -0.0044]], + dtype=torch.bfloat16), 'net.blocks.22.mlp.layer2.weight': tensor([[ 0.0126, -0.0034, -0.0110, ..., 0.0099, 0.0159, -0.0125], + [-0.0054, 0.0129, 0.0048, ..., 0.0041, -0.0142, 0.0166], + [-0.0056, 0.0106, -0.0194, ..., 0.0193, 0.0064, 0.0145], + ..., + [ 0.0242, 0.0178, -0.0131, ..., 0.0156, 0.0031, 0.0151], + [-0.0118, 0.0031, -0.0084, ..., -0.0093, -0.0011, -0.0010], + [ 0.0035, 0.0032, 0.0243, ..., -0.0114, -0.0028, 0.0053]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.1.weight': tensor([[-1.1597e-02, -1.0742e-02, -1.0132e-02, ..., 8.0490e-04, + 8.3923e-04, 7.7820e-04], + [-6.0791e-02, -5.9814e-02, -5.8105e-02, ..., 7.0572e-04, + 7.4005e-04, 7.7438e-04], + [ 5.0049e-02, 5.0293e-02, 5.0537e-02, ..., -5.0068e-05, + -6.4850e-05, 1.0824e-04], + ..., + [ 4.0771e-02, 4.0283e-02, 3.8818e-02, ..., -8.9264e-04, + -8.9264e-04, -9.0408e-04], + [-5.3223e-02, -5.2734e-02, -5.1514e-02, ..., 5.6458e-04, + 6.1417e-04, 6.0654e-04], + [ 7.2754e-02, 7.1777e-02, 7.0801e-02, ..., -3.9673e-04, + -5.1880e-04, -4.8065e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_self_attn.2.weight': tensor([[-5.0049e-03, 5.2185e-03, -4.2419e-03, ..., 4.2419e-03, + 4.2725e-03, -5.2185e-03], + [ 2.8076e-03, 1.5564e-03, 2.2583e-03, ..., -3.8757e-03, + 3.9368e-03, 2.0218e-04], + [ 4.9210e-04, 2.6703e-03, 2.0266e-05, ..., -5.9128e-04, + 2.2278e-03, -1.6479e-03], + ..., + [-2.4658e-02, 3.6621e-02, -3.1128e-02, ..., 1.4648e-02, + 1.4404e-02, -3.7109e-02], + [-3.0029e-02, 2.3071e-02, -3.2959e-02, ..., 2.5024e-02, + -6.0425e-03, -2.9297e-02], + [ 3.1738e-02, -2.2827e-02, 3.4180e-02, ..., -2.4902e-02, + 9.8419e-04, 2.8564e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.1.weight': tensor([[-5.4932e-03, -4.6082e-03, -3.9978e-03, ..., -1.7834e-04, + -2.2793e-04, -2.7275e-04], + [-4.4678e-02, -4.3213e-02, -4.1748e-02, ..., 2.5368e-04, + 2.4605e-04, 1.6117e-04], + [ 3.9307e-02, 3.7598e-02, 3.7598e-02, ..., -2.2221e-04, + -1.4400e-04, -1.0872e-04], + ..., + [ 6.4941e-02, 6.3965e-02, 6.2256e-02, ..., 1.4019e-04, + 2.3723e-05, 1.7357e-04], + [ 4.4922e-02, 4.4434e-02, 4.3457e-02, ..., -2.1362e-04, + -3.0136e-04, -2.1648e-04], + [-1.0559e-02, -9.8877e-03, -9.3384e-03, ..., 1.7738e-04, + 2.3079e-04, 1.8883e-04]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_cross_attn.2.weight': tensor([[-3.1891e-03, 4.9744e-03, -5.0049e-03, ..., 9.4223e-04, + -5.0659e-03, 3.0365e-03], + [-8.4400e-05, -2.6894e-04, 2.4986e-04, ..., -2.5024e-03, + 5.5313e-04, 3.6001e-05], + [-6.9275e-03, 6.2866e-03, -6.5613e-03, ..., 5.1880e-03, + -6.3477e-03, 4.1199e-03], + ..., + [-2.0508e-02, 1.8555e-02, -1.9287e-02, ..., 2.8564e-02, + -1.8188e-02, 3.0518e-02], + [ 2.3315e-02, -1.9165e-02, 2.0020e-02, ..., -2.1362e-02, + 1.8433e-02, -2.1118e-02], + [-3.2471e-02, 2.8442e-02, -2.9297e-02, ..., 3.9795e-02, + -2.7710e-02, 3.8818e-02]], dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.1.weight': tensor([[-0.0491, -0.0481, -0.0479, ..., -0.0003, -0.0002, -0.0002], + [ 0.1865, 0.1826, 0.1797, ..., -0.0002, -0.0002, -0.0002], + [-0.1221, -0.1187, -0.1162, ..., 0.0008, 0.0008, 0.0007], + ..., + [-0.0016, -0.0016, -0.0017, ..., -0.0004, -0.0003, -0.0004], + [ 0.0688, 0.0669, 0.0654, ..., -0.0014, -0.0014, -0.0014], + [-0.0145, -0.0142, -0.0146, ..., -0.0002, -0.0003, -0.0003]], + dtype=torch.bfloat16), 'net.blocks.22.adaln_modulation_mlp.2.weight': tensor([[ 6.1512e-05, -1.1139e-03, -1.0834e-03, ..., -2.8229e-04, + 2.1973e-03, -2.5558e-04], + [-4.4632e-04, 1.4877e-03, -4.2343e-04, ..., -2.3270e-04, + 8.6594e-04, -2.9945e-04], + [-1.2875e-04, 8.7357e-04, -2.7771e-03, ..., 2.7084e-04, + 5.9509e-03, 6.0558e-05], + ..., + [ 5.5420e-02, -7.1777e-02, 7.3730e-02, ..., 4.7119e-02, + -4.5166e-02, 4.9316e-02], + [ 5.0049e-02, -5.7129e-02, 6.3965e-02, ..., 4.7119e-02, + -4.2725e-02, 4.8584e-02], + [-5.7373e-02, 7.2266e-02, -8.4961e-02, ..., -5.0781e-02, + 7.3730e-02, -5.2979e-02]], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_proj.weight': tensor([[ 0.0045, 0.0008, -0.0023, ..., -0.0039, -0.0020, 0.0016], + [-0.0014, 0.0067, 0.0166, ..., -0.0019, 0.0042, -0.0139], + [-0.0045, 0.0047, -0.0121, ..., 0.0087, 0.0102, -0.0031], + ..., + [-0.0021, -0.0087, -0.0114, ..., -0.0100, -0.0266, 0.0041], + [ 0.0021, 0.0176, 0.0057, ..., 0.0030, 0.0115, 0.0034], + [ 0.0141, 0.0097, 0.0189, ..., -0.0006, 0.0115, -0.0154]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm.weight': tensor([0.6953, 1.9531, 1.0234, 0.6133, 0.8164, 1.0547, 1.4453, 1.5625, 1.2734, + 1.5234, 1.5156, 0.5820, 0.4648, 1.3906, 1.4609, 1.2734, 1.2734, 1.4141, + 1.3047, 1.2656, 1.4297, 1.5000, 1.3359, 1.0000, 0.1895, 1.0703, 1.0078, + 0.8945, 1.4922, 1.1172, 1.5391, 1.4766, 0.6836, 1.2969, 0.2617, 1.3516, + 1.3672, 1.2578, 1.5625, 1.2969, 0.8320, 0.7695, 1.2344, 1.9688, 1.8672, + 1.7500, 0.9453, 0.7500, 1.1953, 0.7617, 0.7227, 1.1172, 1.4453, 1.3125, + 0.9492, 1.4297, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.9219, 0.6953, 1.1562, 0.6445, 0.3379, 1.5625, 0.0967, 1.3047, + 0.8984, 0.6016, 1.3906, 1.4219, 0.7969, 1.1875, 1.2344, 1.2656, 0.7305, + 1.4297, 1.4141, 1.1719, 1.4688, 0.8984, 1.6562, 2.1250, 0.6523, 2.2969, + 1.1328, 0.8906, 1.7031, 1.2656, 1.1328, 1.3281, 1.4141, 1.2344, 1.3047, + 1.1484, 0.4082, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 0.6758, + 0.8398, 1.6172, 2.3438, 0.8047, 1.9766, 0.9141, 0.8867, 1.4531, 1.4531, + 1.2578, 1.0000, 1.1406, 0.3789, 0.9102, 1.5078, 0.8008, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.k_proj.weight': tensor([[-0.0142, -0.0016, 0.0014, ..., -0.0248, -0.0080, -0.0124], + [-0.0044, 0.0288, -0.0053, ..., 0.0102, 0.0197, -0.0178], + [ 0.0113, 0.0081, -0.0003, ..., 0.0044, -0.0118, 0.0037], + ..., + [-0.0037, -0.0179, -0.0016, ..., -0.0159, 0.0095, 0.0042], + [ 0.0066, -0.0081, -0.0099, ..., -0.0103, 0.0031, -0.0022], + [-0.0165, -0.0067, 0.0037, ..., -0.0232, -0.0234, -0.0074]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm.weight': tensor([1.4062, 1.5859, 0.7773, 0.5352, 0.7734, 1.1406, 1.4375, 1.5391, 1.2266, + 1.5312, 1.4844, 0.5859, 0.4551, 1.3828, 1.4531, 1.2656, 1.2500, 1.4062, + 1.3125, 1.2656, 1.4297, 1.5000, 1.6562, 1.4375, 0.1357, 1.3906, 0.7578, + 0.9219, 1.5078, 1.1562, 1.5234, 1.4844, 0.6875, 1.2812, 0.2598, 1.3516, + 1.3672, 1.2578, 1.5703, 1.2969, 0.8320, 0.7695, 1.2344, 1.4688, 1.5234, + 1.5078, 1.2500, 0.7500, 1.2812, 0.7578, 0.6641, 1.1406, 1.4375, 1.3203, + 0.9414, 1.4219, 1.0703, 1.3281, 1.2188, 1.4062, 1.3828, 1.4609, 1.2812, + 1.1484, 1.4297, 0.9102, 1.2812, 0.6602, 0.4688, 1.4766, 0.0510, 1.3438, + 0.9023, 0.5898, 1.4375, 1.4219, 0.7969, 1.1953, 1.2344, 1.2656, 0.7266, + 1.4375, 1.4141, 1.1719, 1.4688, 0.9023, 1.7109, 1.6016, 0.6484, 2.0156, + 1.1562, 0.8398, 1.7266, 1.2812, 1.1562, 1.3359, 1.3906, 1.2344, 1.3125, + 1.1484, 0.4102, 0.9297, 0.9375, 0.7266, 1.3516, 0.9258, 1.3281, 1.4219, + 1.3594, 1.7031, 1.6797, 0.6953, 1.8672, 0.9414, 0.8906, 1.4141, 1.4531, + 1.2656, 1.0000, 1.1406, 0.3828, 0.9062, 1.5078, 0.8047, 1.0234, 1.3125, + 1.0078, 1.4609], dtype=torch.bfloat16), 'net.blocks.23.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.self_attn.v_proj.weight': tensor([[ 0.0069, -0.0118, 0.0008, ..., 0.0118, 0.0041, 0.0045], + [-0.0054, 0.0239, 0.0105, ..., 0.0204, -0.0198, 0.0108], + [-0.0095, 0.0069, 0.0126, ..., -0.0203, 0.0198, -0.0238], + ..., + [-0.0107, -0.0020, 0.0264, ..., -0.0112, -0.0044, 0.0004], + [ 0.0126, -0.0093, 0.0214, ..., -0.0050, -0.0054, 0.0041], + [ 0.0019, 0.0043, -0.0150, ..., 0.0054, -0.0049, -0.0248]], + dtype=torch.bfloat16), 'net.blocks.23.self_attn.output_proj.weight': tensor([[-3.6163e-03, 5.4932e-04, 5.5237e-03, ..., 4.1504e-03, + -8.1787e-03, -3.5553e-03], + [ 1.8433e-02, 4.2419e-03, 7.4158e-03, ..., -5.8289e-03, + -1.2268e-02, -8.9111e-03], + [-8.3008e-03, 2.3346e-03, 1.3550e-02, ..., -3.8605e-03, + 4.2114e-03, -7.5684e-03], + ..., + [-1.7471e-03, 1.0315e-02, -1.0620e-02, ..., 1.0803e-02, + 5.7068e-03, -8.2850e-06], + [ 8.8501e-03, 1.4160e-02, -5.1270e-03, ..., 9.0332e-03, + -1.2024e-02, -8.8501e-03], + [-4.6015e-05, -6.2561e-03, 2.0447e-03, ..., 4.1199e-03, + -1.0803e-02, 1.7334e-02]], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_proj.weight': tensor([[ 0.0192, -0.0141, 0.0022, ..., 0.0040, -0.0195, 0.0051], + [ 0.0166, -0.0029, -0.0151, ..., -0.0192, 0.0114, 0.0016], + [-0.0187, -0.0130, -0.0018, ..., -0.0128, -0.0038, 0.0141], + ..., + [ 0.0062, -0.0095, -0.0267, ..., 0.0038, -0.0165, -0.0510], + [-0.0017, -0.0178, -0.0040, ..., 0.0057, 0.0151, 0.0178], + [ 0.0181, 0.0068, -0.0121, ..., 0.0132, -0.0199, -0.0089]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm.weight': tensor([ 3.5889e-02, -3.3760e-04, 7.9346e-04, 8.8501e-03, 5.9204e-03, + 4.4922e-01, 1.1778e-04, -1.6594e-04, 5.1880e-03, 8.0566e-03, + 2.6245e-02, 9.8228e-05, 1.3828e-04, 9.2506e-05, -6.5327e-05, + 7.6294e-05, 7.2327e-03, 5.6624e-06, 1.2695e-02, 2.4719e-03, + 2.9755e-04, -2.1362e-04, 4.1199e-03, 1.2875e-04, -1.2589e-04, + 7.0190e-04, 9.4604e-03, 5.9509e-03, 3.3722e-03, -1.3638e-04, + 2.8687e-03, 7.2021e-03, 1.9531e-02, 2.1210e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.4343e-03, 1.0681e-02, -5.9366e-05, + 1.3428e-03, 6.3171e-03, -3.4637e-03, 7.6599e-03, -1.9455e-04, + -2.2125e-04, -1.4877e-04, 3.4714e-04, 2.5787e-03, 1.9989e-03, + 1.2146e-02, 8.6060e-03, 5.8899e-03, 2.0862e-05, 2.7771e-03, + -1.6975e-04, 2.0123e-04, -7.5340e-05, 1.4648e-02, 1.9932e-04, + 3.8281e-01, 3.1128e-02, 2.0386e-02, 1.2112e-04, -8.9645e-05, + -1.6785e-04, 3.5156e-01, 4.6082e-03, -2.4557e-05, 1.7773e-01, + 2.2697e-04, 1.3000e-02, 6.7234e-05, -1.6785e-04, -1.3161e-04, + 5.5542e-03, 3.5763e-05, 5.5847e-03, 9.0942e-03, 3.0365e-03, + 6.6757e-05, 3.5477e-04, 1.3199e-03, 5.7031e-01, 4.7363e-02, + 1.7090e-03, 9.7046e-03, 2.1210e-03, 9.8267e-03, -8.5831e-06, + -2.7847e-04, -1.0586e-04, -1.0347e-04, 8.1635e-04, 7.6294e-05, + -3.3569e-04, 2.1648e-04, -2.7084e-04, 2.1648e-04, 1.8164e-01, + 1.6308e-04, -1.9360e-04, 2.3651e-04, -8.1539e-05, 6.6833e-03, + -7.7724e-05, 4.1504e-03, -1.2875e-04, 2.7466e-03, 5.2185e-03, + 1.5736e-04, 1.2283e-03, 1.5869e-02, 7.6294e-04, 4.3335e-03, + -1.7262e-04, 7.2327e-03, 1.4404e-02, 5.7983e-03, -5.9509e-03, + 2.3746e-04, 7.5150e-04, 1.8457e-01, 7.1049e-05, 2.0508e-01, + 2.3651e-04, 8.8501e-03, 1.6602e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_proj.weight': tensor([[-0.0034, 0.0005, 0.0048, ..., -0.0049, 0.0027, 0.0009], + [-0.0087, -0.0015, -0.0002, ..., 0.0035, -0.0045, -0.0084], + [-0.0134, -0.0015, 0.0009, ..., -0.0018, -0.0013, -0.0063], + ..., + [-0.0031, -0.0080, -0.0061, ..., -0.0020, -0.0093, 0.0055], + [-0.0024, 0.0059, 0.0023, ..., 0.0036, 0.0014, -0.0007], + [-0.0044, 0.0091, -0.0046, ..., -0.0020, 0.0012, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm.weight': tensor([ 3.5889e-02, -3.4142e-04, -7.7057e-04, 8.8501e-03, 5.9814e-03, + 4.4531e-01, 1.1873e-04, 1.6880e-04, 5.0964e-03, 7.8735e-03, + 2.6001e-02, 9.7752e-05, -1.4019e-04, -8.9169e-05, 6.8188e-05, + 6.8188e-05, 7.2937e-03, -6.5565e-06, 1.2451e-02, 2.4872e-03, + 2.9755e-04, -2.1267e-04, 4.1809e-03, -1.3065e-04, -1.2589e-04, + -7.2479e-04, 9.7046e-03, 5.4932e-03, 3.3875e-03, 1.3733e-04, + 2.7771e-03, 9.2163e-03, 1.9531e-02, 2.0752e-03, 1.2158e-01, + 3.1641e-01, 1.7166e-04, 1.5793e-03, 1.0315e-02, -6.7234e-05, + 1.3275e-03, 6.3782e-03, 3.3112e-03, 7.7209e-03, -1.9646e-04, + -2.2030e-04, -1.4782e-04, 3.2806e-04, 2.7618e-03, 1.9302e-03, + -8.9722e-03, 8.6060e-03, 5.7373e-03, 2.4080e-05, 2.5787e-03, + -1.7357e-04, 2.0123e-04, -1.0252e-04, 1.4893e-02, 2.0027e-04, + 3.8281e-01, 3.0396e-02, 2.0142e-02, -1.2159e-04, -2.3961e-05, + -1.6689e-04, 3.4766e-01, 4.6082e-03, -3.0518e-05, 1.7773e-01, + -2.2411e-04, 1.3123e-02, 6.6757e-05, -1.6785e-04, -1.2970e-04, + 5.7983e-03, 3.0756e-05, 5.6152e-03, 8.9722e-03, 3.2959e-03, + -1.8692e-04, 3.4904e-04, 5.4169e-04, 5.7422e-01, 4.7363e-02, + 1.8539e-03, 9.2163e-03, 2.1667e-03, 9.8877e-03, -1.0967e-05, + -2.7657e-04, -1.0538e-04, -1.0395e-04, 7.8583e-04, -7.6294e-05, + -3.3569e-04, 2.1553e-04, -2.7084e-04, -2.1648e-04, 1.8066e-01, + 1.6212e-04, -1.9932e-04, -2.3174e-04, -7.8201e-05, 6.5918e-03, + -7.8201e-05, 4.2419e-03, -1.2779e-04, 2.7466e-03, 5.0049e-03, + -1.5545e-04, 2.0294e-03, 1.6235e-02, 7.7057e-04, -3.5553e-03, + -1.7357e-04, 7.2937e-03, 1.4526e-02, 5.7983e-03, 5.9509e-03, + 2.3842e-04, 7.4005e-04, 1.8457e-01, 6.7711e-05, 2.0410e-01, + -2.3651e-04, 8.7280e-03, 1.6479e-02], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.v_proj.weight': tensor([[ 0.0176, -0.0026, 0.0106, ..., 0.0007, -0.0006, 0.0099], + [ 0.0058, -0.0033, 0.0215, ..., 0.0044, -0.0048, 0.0094], + [ 0.0001, 0.0068, 0.0137, ..., 0.0046, 0.0153, -0.0078], + ..., + [-0.0103, -0.0044, -0.0199, ..., -0.0062, -0.0248, -0.0277], + [ 0.0188, 0.0067, 0.0273, ..., 0.0070, 0.0128, 0.0206], + [-0.0088, -0.0020, -0.0129, ..., 0.0052, 0.0162, 0.0076]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.output_proj.weight': tensor([[ 0.0037, 0.0065, 0.0028, ..., 0.0110, -0.0021, -0.0096], + [-0.0085, 0.0024, 0.0013, ..., -0.0347, 0.0043, 0.0216], + [-0.0015, -0.0018, -0.0056, ..., -0.0195, -0.0034, 0.0198], + ..., + [-0.0052, 0.0003, -0.0048, ..., -0.0157, 0.0081, -0.0019], + [ 0.0024, 0.0070, 0.0010, ..., 0.0003, 0.0117, -0.0181], + [ 0.0120, -0.0019, 0.0154, ..., -0.0145, 0.0172, -0.0079]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img.weight': tensor([[-0.0123, -0.0297, 0.0049, ..., 0.0347, 0.0334, 0.0315], + [ 0.0231, 0.0142, -0.0104, ..., 0.0165, -0.0280, 0.0481], + [-0.0090, -0.0035, -0.0027, ..., 0.0287, 0.0273, 0.0601], + ..., + [-0.0222, 0.0127, 0.0603, ..., 0.0121, 0.0187, 0.0260], + [ 0.0366, -0.0099, -0.0049, ..., 0.0302, -0.0018, -0.0297], + [ 0.0146, -0.0153, -0.0265, ..., -0.0141, 0.0023, -0.0036]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.v_img.weight': tensor([[ 0.0206, -0.0103, 0.0286, ..., -0.0339, 0.0165, -0.0119], + [-0.0095, 0.0028, -0.0530, ..., -0.0283, 0.0469, -0.0267], + [-0.0186, 0.0508, -0.0469, ..., 0.0037, -0.0222, 0.0262], + ..., + [-0.0147, 0.0043, 0.0260, ..., -0.0042, 0.0120, 0.0305], + [-0.0172, 0.0410, -0.0258, ..., -0.0327, -0.0066, 0.0090], + [ 0.0221, 0.0136, 0.0386, ..., -0.0035, -0.0175, 0.0182]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img.weight': tensor([[ 0.0198, 0.0109, -0.0356, ..., -0.0145, -0.0249, 0.0112], + [-0.0278, 0.0041, -0.0374, ..., 0.0276, -0.0003, 0.0151], + [ 0.0089, 0.0237, -0.0039, ..., 0.0527, -0.0031, -0.0437], + ..., + [-0.0398, 0.0317, -0.0173, ..., -0.0244, -0.0072, -0.0095], + [-0.0067, 0.0264, -0.0097, ..., -0.0108, -0.0603, 0.0134], + [-0.0148, 0.0256, 0.0130, ..., 0.0135, 0.0229, 0.0081]], + dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.23.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.23.mlp.layer1.weight': tensor([[-0.0291, -0.0154, -0.0059, ..., -0.0100, 0.0012, -0.0076], + [ 0.0327, 0.0190, -0.0147, ..., -0.0051, -0.0077, 0.0312], + [ 0.0004, -0.0099, -0.0237, ..., 0.0077, 0.0063, -0.0256], + ..., + [-0.0108, 0.0112, 0.0177, ..., -0.0038, -0.0021, 0.0129], + [-0.0048, -0.0107, 0.0056, ..., 0.0078, 0.0232, 0.0297], + [-0.0085, -0.0041, 0.0018, ..., -0.0037, -0.0052, 0.0134]], + dtype=torch.bfloat16), 'net.blocks.23.mlp.layer2.weight': tensor([[-3.4668e-02, 9.8267e-03, -2.3041e-03, ..., -3.5400e-03, + 8.8501e-03, 1.5625e-02], + [ 3.5156e-02, -1.0132e-02, 5.1575e-03, ..., -1.8066e-02, + 1.8082e-03, 4.8218e-03], + [-3.0518e-03, -5.8899e-03, 1.4893e-02, ..., -1.0071e-02, + 9.5215e-03, 4.8828e-03], + ..., + [-1.0498e-02, -1.6113e-02, -5.9814e-03, ..., 9.4604e-03, + -1.8997e-03, -4.3030e-03], + [ 7.1411e-03, -1.6113e-02, 2.3346e-03, ..., -7.0190e-03, + 6.9275e-03, 2.4719e-03], + [-2.0752e-03, -1.8239e-05, 3.8605e-03, ..., 3.7231e-03, + -1.0254e-02, 1.2634e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.1.weight': tensor([[-8.9355e-02, -8.6914e-02, -8.6914e-02, ..., 3.8147e-04, + 2.6321e-04, 4.3297e-04], + [ 3.6865e-02, 3.6133e-02, 3.5156e-02, ..., -2.9755e-03, + -2.9907e-03, -2.9449e-03], + [-9.4238e-02, -9.2773e-02, -9.1797e-02, ..., 2.9945e-04, + 2.7466e-04, 3.2043e-04], + ..., + [-3.8330e-02, -3.8818e-02, -3.9062e-02, ..., -7.3242e-04, + -6.9046e-04, -6.2943e-04], + [ 9.2773e-02, 9.1797e-02, 9.0820e-02, ..., -1.2684e-04, + -2.2507e-04, -9.0599e-05], + [ 4.2480e-02, 4.1504e-02, 4.0527e-02, ..., -1.8768e-03, + -1.9836e-03, -1.9150e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_self_attn.2.weight': tensor([[ 3.6774e-03, -8.7891e-03, 4.6082e-03, ..., 3.0975e-03, + -4.5471e-03, -1.0620e-02], + [-1.9302e-03, -4.2419e-03, -1.3885e-03, ..., -1.8234e-03, + 1.9226e-03, -4.3640e-03], + [ 1.6594e-04, -3.3417e-03, 1.1826e-04, ..., -3.8910e-04, + 1.6332e-05, -3.0060e-03], + ..., + [-3.4180e-02, 2.7710e-02, -3.6377e-02, ..., -2.6489e-02, + 3.4424e-02, 3.3203e-02], + [ 3.0396e-02, 9.3384e-03, 2.9297e-02, ..., 3.0884e-02, + -2.9785e-02, 7.4158e-03], + [-3.3203e-02, 1.0834e-03, -3.1982e-02, ..., -3.2227e-02, + 3.1982e-02, 2.9449e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.1.weight': tensor([[-6.6406e-02, -6.5918e-02, -6.3477e-02, ..., 1.8024e-04, + 2.5558e-04, 1.0586e-04], + [-6.5430e-02, -6.3477e-02, -6.1768e-02, ..., 1.0014e-04, + 1.2684e-04, 8.7261e-05], + [ 6.4941e-02, 6.4453e-02, 6.2500e-02, ..., -2.6894e-04, + -2.2030e-04, -2.1267e-04], + ..., + [-2.1484e-02, -2.0874e-02, -2.0020e-02, ..., -9.4891e-05, + -1.3542e-04, -2.1553e-04], + [ 8.2520e-02, 8.0078e-02, 7.7637e-02, ..., 1.4400e-04, + 1.3447e-04, 1.3065e-04], + [ 3.4180e-02, 3.2471e-02, 3.1982e-02, ..., 1.5640e-04, + 2.8229e-04, 1.6499e-04]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_cross_attn.2.weight': tensor([[-4.3631e-05, -6.5231e-04, 6.1417e-04, ..., -6.1035e-04, + 2.5330e-03, 2.9297e-03], + [ 3.6621e-03, 2.3499e-03, -2.9602e-03, ..., 3.6316e-03, + -1.9150e-03, -4.9133e-03], + [-1.8845e-03, -1.3962e-03, 1.6174e-03, ..., -2.6093e-03, + 1.4954e-03, 4.6997e-03], + ..., + [-3.5858e-03, -3.1891e-03, 2.9144e-03, ..., 2.7100e-02, + -8.4229e-03, 3.9551e-02], + [-7.2632e-03, -4.1809e-03, 5.6152e-03, ..., 2.2705e-02, + -2.2125e-03, 3.4912e-02], + [ 4.6082e-03, 1.3428e-03, -3.2806e-03, ..., -2.5635e-02, + 9.5215e-03, -3.4424e-02]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.1.weight': tensor([[ 8.9844e-02, 8.7891e-02, 8.6914e-02, ..., -5.3644e-06, + -6.4373e-05, -4.1723e-06], + [ 9.7168e-02, 9.6191e-02, 9.4727e-02, ..., -5.4121e-05, + -6.3419e-05, -1.3447e-04], + [ 8.3984e-02, 8.2520e-02, 8.1055e-02, ..., -9.8419e-04, + -9.9182e-04, -8.8120e-04], + ..., + [-6.2500e-02, -6.1279e-02, -5.9570e-02, ..., 1.3199e-03, + 1.3885e-03, 1.1978e-03], + [ 6.7871e-02, 6.6406e-02, 6.5430e-02, ..., -1.3504e-03, + -1.4725e-03, -1.4343e-03], + [-6.9824e-02, -6.7383e-02, -6.6406e-02, ..., 1.5945e-03, + 1.6174e-03, 1.5182e-03]], dtype=torch.bfloat16), 'net.blocks.23.adaln_modulation_mlp.2.weight': tensor([[-0.0008, -0.0008, 0.0066, ..., -0.0125, 0.0117, -0.0101], + [ 0.0004, 0.0004, 0.0005, ..., -0.0013, 0.0017, -0.0007], + [-0.0003, -0.0003, 0.0020, ..., 0.0022, -0.0007, -0.0018], + ..., + [-0.0640, -0.0664, -0.1514, ..., 0.1670, -0.1865, 0.1797], + [-0.0649, -0.0664, -0.1123, ..., 0.0913, -0.1167, 0.1211], + [ 0.0723, 0.0747, 0.1279, ..., -0.1250, 0.1445, -0.1396]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_proj.weight': tensor([[-0.0334, -0.0119, 0.0036, ..., -0.0124, 0.0101, 0.0064], + [-0.0115, -0.0018, -0.0148, ..., -0.0040, -0.0166, -0.0038], + [ 0.0095, 0.0036, -0.0090, ..., 0.0188, -0.0161, 0.0038], + ..., + [-0.0025, -0.0021, 0.0082, ..., 0.0354, -0.0100, 0.0192], + [-0.0312, 0.0049, 0.0050, ..., 0.0071, 0.0186, -0.0148], + [-0.0120, -0.0019, 0.0133, ..., -0.0023, -0.0078, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm.weight': tensor([1.3359, 1.7344, 1.1094, 1.1641, 0.7500, 0.9414, 0.6797, 0.9180, 1.3828, + 1.4844, 1.0781, 1.0781, 1.0078, 1.4766, 0.9141, 1.0156, 1.2812, 1.4375, + 1.3672, 1.0156, 1.3438, 1.4375, 0.7188, 0.8398, 1.7500, 0.7969, 0.0383, + 1.3672, 1.5781, 1.4844, 1.4219, 1.3359, 1.3984, 1.4688, 1.4297, 1.0703, + 1.1328, 1.3828, 1.4062, 1.0469, 1.1250, 1.1797, 1.2500, 1.5469, 0.7344, + 0.8477, 0.3887, 1.8438, 1.1016, 0.6328, 1.4375, 0.4531, 0.9727, 1.1484, + 1.3359, 1.3750, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.4141, 0.7383, 0.6562, 1.0938, 1.0625, 1.4219, 0.9414, 0.9531, + 1.0000, 0.8320, 1.5156, 1.3672, 1.2109, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.9297, 0.9102, 0.9258, 0.7891, + 0.2246, 1.5234, 1.0703, 1.1641, 1.2891, 1.3281, 1.3984, 1.1250, 1.4375, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4219, 1.3359, 1.3906, 1.2266, 0.8555, + 1.6250, 1.6875, 0.6289, 0.8945, 1.3047, 1.5859, 0.6914, 1.2344, 1.1094, + 1.1094, 1.1250, 1.3047, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.k_proj.weight': tensor([[-0.0618, 0.0044, -0.0048, ..., 0.0090, 0.0160, 0.0057], + [ 0.0415, -0.0040, -0.0040, ..., -0.0046, -0.0027, -0.0153], + [-0.0439, -0.0041, 0.0195, ..., 0.0086, 0.0057, -0.0038], + ..., + [ 0.0036, 0.0063, 0.0099, ..., 0.0079, -0.0082, 0.0091], + [ 0.0096, -0.0042, 0.0131, ..., 0.0020, 0.0049, -0.0038], + [-0.0100, 0.0058, -0.0135, ..., -0.0195, 0.0054, -0.0062]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm.weight': tensor([1.6172, 1.5781, 0.9688, 1.1953, 0.6680, 1.1484, 0.6172, 0.9180, 1.3750, + 1.4844, 1.1250, 1.1016, 1.0078, 1.4766, 0.9141, 1.0078, 1.2812, 1.4297, + 1.3672, 1.0156, 1.3438, 1.4375, 1.4219, 0.9492, 1.4844, 0.7148, 0.0752, + 1.4453, 1.5469, 1.4531, 1.3906, 1.3359, 1.4062, 1.4609, 1.4297, 1.0781, + 1.1406, 1.3828, 1.4062, 1.0469, 1.1328, 1.1797, 1.2500, 1.2266, 1.1875, + 1.1562, 0.4316, 1.6719, 1.2266, 0.7578, 1.4453, 0.4609, 0.9609, 1.1484, + 1.3281, 1.3672, 1.0156, 1.4219, 1.4453, 1.4375, 1.1797, 1.3203, 1.3594, + 1.4062, 1.2188, 1.2734, 1.1016, 0.9727, 1.1172, 1.3672, 0.9883, 0.9531, + 1.0078, 0.8789, 1.4922, 1.3594, 1.2031, 1.1094, 1.5469, 1.4375, 1.1797, + 0.8789, 1.3672, 1.4609, 1.2891, 1.4375, 1.5156, 0.6797, 1.1875, 0.8242, + 0.2207, 1.4609, 1.3047, 1.2266, 1.3203, 1.3281, 1.3984, 1.1328, 1.4297, + 1.5078, 1.4688, 1.1406, 1.2656, 1.4297, 1.3359, 1.3906, 1.2266, 1.4688, + 0.7109, 1.5781, 0.6055, 1.1641, 1.2109, 1.5469, 0.6367, 1.2344, 1.1172, + 1.1094, 1.1250, 1.3203, 1.0781, 1.3672, 1.0547, 1.3516, 1.2969, 1.1094, + 0.8984, 1.0312], dtype=torch.bfloat16), 'net.blocks.24.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.self_attn.v_proj.weight': tensor([[ 0.0001, -0.0104, 0.0052, ..., -0.0025, -0.0157, -0.0081], + [ 0.0031, -0.0035, -0.0005, ..., -0.0042, -0.0022, -0.0033], + [-0.0055, 0.0187, 0.0043, ..., -0.0095, -0.0165, -0.0033], + ..., + [-0.0098, 0.0009, 0.0098, ..., -0.0232, -0.0194, -0.0005], + [ 0.0079, 0.0269, 0.0026, ..., 0.0050, 0.0160, -0.0217], + [-0.0064, -0.0049, -0.0063, ..., -0.0002, 0.0051, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.24.self_attn.output_proj.weight': tensor([[-1.0681e-03, 4.7607e-03, 1.5198e-02, ..., -1.1230e-02, + -4.6997e-03, -1.2085e-02], + [ 1.9287e-02, 1.5137e-02, 1.6022e-03, ..., 3.5889e-02, + 1.1978e-03, -9.4604e-04], + [-1.3672e-02, -4.3640e-03, 1.5564e-02, ..., 1.0824e-04, + -2.5330e-03, -7.6904e-03], + ..., + [ 7.9727e-04, -1.6846e-02, 1.2024e-02, ..., -1.0864e-02, + -6.9809e-04, 1.1047e-02], + [-4.7607e-03, 1.9531e-02, -2.0508e-02, ..., -3.3875e-03, + -8.1787e-03, -1.0529e-03], + [ 2.6733e-02, -1.5137e-02, 6.1340e-03, ..., -9.8267e-03, + 1.7334e-02, 4.6253e-05]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_proj.weight': tensor([[-1.4782e-04, -5.9204e-03, -1.7578e-02, ..., -6.7139e-03, + -2.3560e-02, -2.3651e-03], + [ 1.9150e-03, 5.6152e-03, -1.2146e-02, ..., -6.4697e-03, + -2.2217e-02, 2.4902e-02], + [-6.9275e-03, -5.7373e-03, -6.8665e-04, ..., 8.2397e-03, + -1.6235e-02, -6.9275e-03], + ..., + [ 3.1982e-02, -5.3406e-03, -1.2634e-02, ..., 4.9438e-03, + 1.0315e-02, -8.9111e-03], + [-3.7994e-03, 6.5613e-03, 1.1597e-02, ..., 8.7280e-03, + -7.0095e-05, -1.0910e-03], + [-1.3611e-02, 2.5024e-03, -8.7280e-03, ..., -1.9775e-02, + -2.7222e-02, -1.5137e-02]], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm.weight': tensor([ 1.4019e-04, -1.3638e-04, 4.5776e-03, -1.1139e-03, 1.1265e-05, + 3.7909e-05, 9.8419e-04, -3.1471e-05, -3.1233e-05, -1.2398e-04, + 1.7944e-02, 4.4531e-01, 3.7537e-03, -1.3580e-03, 1.0872e-04, + 3.7384e-04, -2.2125e-04, 6.8054e-03, 1.1158e-04, -3.3855e-05, + 2.5558e-04, 3.3855e-05, 2.1875e-01, 7.3242e-03, -7.5817e-05, + -2.3246e-05, 2.2736e-03, -1.4687e-04, 5.2691e-05, 5.5695e-04, + -1.4400e-04, 1.4305e-04, 2.8253e-05, 7.7486e-06, 1.3733e-04, + 8.7738e-05, 1.1253e-04, 2.4796e-04, -1.2016e-04, -1.6403e-03, + -5.3167e-05, 1.3256e-04, 1.9073e-04, -6.4373e-05, 1.6895e-01, + 7.7724e-05, -1.6117e-04, 9.6321e-05, 4.8584e-02, -9.0027e-04, + 1.5442e-02, 5.2261e-04, 9.3937e-05, -3.4094e-05, 1.2398e-04, + 9.2030e-05, -1.4305e-04, 1.3447e-04, 2.0504e-04, -1.4973e-04, + 6.8665e-04, -8.5354e-05, -1.5736e-04, -6.7711e-05, 2.6321e-04, + -1.9550e-05, 3.2959e-03, -1.0312e-05, -9.9838e-07, 2.3484e-05, + 1.2493e-04, -2.8491e-05, 1.4782e-04, -5.0306e-05, -1.2302e-04, + -4.6492e-05, 2.5586e-01, 7.8678e-05, 8.3618e-03, 7.2002e-05, + -8.6308e-05, -8.8215e-05, -3.6621e-04, 1.6570e-05, 1.0910e-03, + -1.2302e-04, -1.4973e-04, -1.6880e-04, -1.3292e-05, 2.3340e-01, + -1.2398e-04, 1.4973e-04, 1.6403e-04, 7.5989e-03, -1.9646e-04, + 2.5034e-05, 2.8133e-05, 2.1973e-03, 3.1853e-04, 3.0136e-04, + -3.9864e-04, 1.3733e-04, 1.6689e-04, -4.6968e-05, 6.7711e-05, + 1.3924e-04, -4.6539e-04, 1.4687e-04, 1.6113e-02, 3.7695e-01, + -1.6785e-04, 8.9111e-03, 5.9605e-06, 2.1011e-06, 3.3008e-01, + -1.6117e-04, 1.8120e-04, 4.6015e-05, 1.2112e-04, 1.5259e-03, + 5.6744e-05, -7.4863e-05, -1.2064e-04, 1.2634e-02, 3.8147e-04, + 1.1587e-04, -1.9908e-05, 3.2187e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_proj.weight': tensor([[-0.0099, -0.0033, -0.0042, ..., -0.0020, 0.0075, 0.0053], + [ 0.0060, 0.0026, -0.0020, ..., 0.0034, -0.0067, -0.0056], + [-0.0056, -0.0026, -0.0043, ..., 0.0002, -0.0054, 0.0009], + ..., + [-0.0047, 0.0071, 0.0067, ..., -0.0030, -0.0049, -0.0023], + [-0.0076, -0.0069, -0.0087, ..., 0.0010, 0.0068, 0.0151], + [-0.0025, -0.0045, -0.0065, ..., -0.0027, 0.0047, -0.0114]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm.weight': tensor([ 1.3924e-04, 1.3828e-04, 1.2207e-02, 1.1215e-03, 9.2387e-06, + -3.5763e-05, 9.6130e-04, 4.4107e-05, -3.1233e-05, 1.2398e-04, + 1.7822e-02, 4.4531e-01, -3.9062e-03, -1.3657e-03, -1.1063e-04, + -3.7193e-04, 2.2125e-04, 6.6528e-03, 1.2684e-04, -4.2915e-05, + 2.5749e-04, -3.0160e-05, 2.1875e-01, 7.3242e-03, 7.6294e-05, + 4.6015e-05, -2.2736e-03, -1.4782e-04, -5.2929e-05, 5.5695e-04, + 1.4400e-04, -1.4400e-04, 2.8253e-05, -6.6459e-06, 1.3916e-02, + 5.4359e-05, -1.1444e-04, -2.4605e-04, -1.2064e-04, -1.6403e-03, + -5.2691e-05, -1.3256e-04, 1.9169e-04, -6.7234e-05, 1.6895e-01, + -9.5844e-05, -1.6212e-04, 1.0252e-04, 4.9561e-02, 8.8882e-04, + 1.5564e-02, 5.0735e-04, 9.2506e-05, 2.3961e-05, 1.2398e-04, + -9.2030e-05, -1.4400e-04, -1.3351e-04, 2.0599e-04, -1.4973e-04, + 5.6076e-04, 5.1737e-05, 1.5831e-04, -5.7936e-05, 2.6703e-04, + 1.2815e-05, 3.5706e-03, -2.0742e-05, -3.2634e-06, 9.5963e-06, + 1.2589e-04, -3.4571e-05, 1.4782e-04, -5.1737e-05, -1.2302e-04, + -3.0994e-05, 2.5195e-01, 8.2493e-05, 8.7280e-03, 8.2970e-05, + 8.3447e-05, 8.6784e-05, 3.0151e-02, -2.2292e-05, -1.0910e-03, + -1.2493e-04, -1.4877e-04, 1.6880e-04, 1.4186e-05, 2.3340e-01, + -1.2302e-04, 1.4973e-04, -1.6308e-04, 7.5989e-03, 1.9836e-04, + -2.3007e-05, 3.0994e-05, 2.2430e-03, 5.3711e-03, -3.0136e-04, + 3.9291e-04, -9.2506e-05, 1.6594e-04, -5.1737e-05, 6.8188e-05, + 1.4114e-04, -4.6730e-04, 1.4687e-04, 1.5076e-02, 3.7695e-01, + 1.6689e-04, 9.2163e-03, -7.6294e-06, 9.1195e-06, 3.2617e-01, + -2.1362e-04, 1.8406e-04, 4.6492e-05, 1.2112e-04, -1.5182e-03, + 5.6744e-05, -7.4387e-05, -1.2207e-04, 1.2024e-02, -3.8338e-04, + 1.1539e-04, -2.0623e-05, 3.3140e-05], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.v_proj.weight': tensor([[ 0.0215, 0.0017, 0.0396, ..., 0.0025, 0.0275, 0.0187], + [-0.0221, -0.0031, -0.0386, ..., -0.0025, -0.0299, -0.0220], + [ 0.0236, 0.0025, 0.0374, ..., 0.0029, 0.0198, 0.0200], + ..., + [-0.0203, -0.0038, -0.0435, ..., -0.0033, -0.0208, -0.0236], + [ 0.0173, 0.0041, 0.0347, ..., 0.0008, 0.0171, 0.0245], + [-0.0233, -0.0039, -0.0403, ..., -0.0038, -0.0223, -0.0239]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.output_proj.weight': tensor([[ 0.0011, 0.0010, -0.0004, ..., -0.0010, -0.0061, -0.0028], + [-0.0117, 0.0135, -0.0139, ..., 0.0127, -0.0126, 0.0142], + [-0.0016, -0.0037, 0.0035, ..., -0.0089, 0.0123, -0.0124], + ..., + [ 0.0125, -0.0003, 0.0083, ..., -0.0073, 0.0026, -0.0042], + [-0.0120, 0.0104, -0.0129, ..., 0.0094, -0.0080, 0.0127], + [-0.0010, 0.0025, -0.0023, ..., 0.0148, -0.0029, 0.0118]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img.weight': tensor([[-0.0195, 0.0036, -0.0128, ..., 0.0063, -0.0371, -0.0064], + [ 0.0064, -0.0130, -0.0028, ..., -0.0204, -0.0013, 0.0057], + [-0.0112, -0.0195, -0.0053, ..., -0.0542, -0.0094, -0.0059], + ..., + [-0.0205, -0.0082, -0.0037, ..., -0.0199, 0.0044, 0.0058], + [-0.0220, -0.0302, -0.0105, ..., 0.0114, 0.0369, 0.0052], + [-0.0117, -0.0160, 0.0496, ..., -0.0435, 0.0134, -0.0253]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.v_img.weight': tensor([[ 0.0017, 0.0347, 0.0095, ..., 0.0277, -0.0208, -0.0220], + [-0.0034, 0.0056, 0.0238, ..., -0.0140, -0.0320, 0.0041], + [-0.0193, 0.0364, -0.0035, ..., -0.0076, 0.0354, 0.0194], + ..., + [ 0.0171, -0.0192, -0.0183, ..., -0.0065, -0.0232, -0.0464], + [-0.0107, -0.0298, -0.0195, ..., 0.0109, 0.0312, 0.0151], + [-0.0074, -0.0116, -0.0195, ..., 0.0034, 0.0107, 0.0109]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img.weight': tensor([[-0.0415, 0.0322, -0.0149, ..., -0.0371, 0.0400, -0.0165], + [ 0.0047, -0.0239, -0.0123, ..., -0.0062, 0.0243, -0.0396], + [-0.0101, 0.0155, -0.0038, ..., -0.0277, 0.0042, 0.0037], + ..., + [ 0.0112, -0.0047, 0.0515, ..., 0.0327, -0.0120, -0.0077], + [-0.0030, 0.0170, 0.0063, ..., -0.0014, 0.0280, -0.0210], + [-0.0016, 0.0076, 0.0055, ..., 0.0153, -0.0044, -0.0184]], + dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.24.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.24.mlp.layer1.weight': tensor([[-0.0089, 0.0142, 0.0155, ..., 0.0041, -0.0004, -0.0170], + [ 0.0017, -0.0069, 0.0090, ..., -0.0187, -0.0045, -0.0045], + [ 0.0143, 0.0050, 0.0129, ..., -0.0060, 0.0043, -0.0156], + ..., + [ 0.0046, 0.0009, -0.0051, ..., 0.0199, -0.0270, 0.0194], + [-0.0042, -0.0250, 0.0110, ..., -0.0044, 0.0053, -0.0315], + [-0.0021, -0.0012, 0.0063, ..., 0.0024, 0.0041, 0.0172]], + dtype=torch.bfloat16), 'net.blocks.24.mlp.layer2.weight': tensor([[-0.0092, 0.0115, -0.0112, ..., -0.0059, -0.0165, -0.0021], + [-0.0134, 0.0114, 0.0093, ..., -0.0094, 0.0226, 0.0121], + [ 0.0004, -0.0004, -0.0203, ..., -0.0186, -0.0134, -0.0089], + ..., + [ 0.0039, -0.0236, 0.0022, ..., -0.0165, -0.0244, 0.0084], + [-0.0212, -0.0071, -0.0048, ..., -0.0154, 0.0126, 0.0008], + [ 0.0197, 0.0150, -0.0111, ..., -0.0151, -0.0044, 0.0037]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.1.weight': tensor([[-0.0527, -0.0515, -0.0503, ..., 0.0007, 0.0009, 0.0009], + [-0.0640, -0.0623, -0.0613, ..., 0.0010, 0.0009, 0.0009], + [-0.0630, -0.0618, -0.0613, ..., 0.0008, 0.0007, 0.0008], + ..., + [-0.0510, -0.0493, -0.0481, ..., 0.0007, 0.0007, 0.0006], + [ 0.0615, 0.0601, 0.0588, ..., -0.0010, -0.0010, -0.0010], + [ 0.0566, 0.0579, 0.0579, ..., 0.0004, 0.0004, 0.0004]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_self_attn.2.weight': tensor([[-9.9945e-04, 1.9989e-03, 3.4943e-03, ..., -1.0681e-03, + -9.5749e-04, -3.6163e-03], + [ 1.8845e-03, -2.0447e-03, -2.7618e-03, ..., 1.8768e-03, + 1.3828e-04, 2.8076e-03], + [-5.8746e-04, -9.1934e-04, -1.3809e-03, ..., -2.7847e-04, + 1.3428e-03, 1.4007e-05], + ..., + [-2.0020e-02, 3.4668e-02, 4.1992e-02, ..., -2.4048e-02, + -1.8921e-03, -4.1748e-02], + [-3.7598e-02, 1.1230e-02, 2.6001e-02, ..., -3.8574e-02, + 2.4170e-02, -4.1504e-02], + [ 2.6367e-02, -2.4292e-02, -3.5889e-02, ..., 3.0029e-02, + -6.5308e-03, 4.2725e-02]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.1.weight': tensor([[ 1.1658e-02, 1.0681e-02, 1.0071e-02, ..., -1.4114e-04, + -2.8801e-04, -1.9073e-04], + [ 7.0312e-02, 6.8359e-02, 6.7383e-02, ..., -1.4877e-04, + -8.2493e-05, -1.8787e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., 1.8024e-04, + 8.7738e-05, 1.0729e-04], + ..., + [-8.3618e-03, -8.3618e-03, -7.9956e-03, ..., 1.4496e-04, + 2.5177e-04, 2.3651e-04], + [ 9.5703e-02, 9.4238e-02, 9.1797e-02, ..., -5.5313e-05, + -1.4496e-04, -1.9741e-04], + [ 7.1777e-02, 7.0801e-02, 6.9336e-02, ..., 8.5831e-05, + 1.7071e-04, 6.7711e-05]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_cross_attn.2.weight': tensor([[-0.0065, -0.0168, -0.0166, ..., 0.0065, -0.0139, -0.0171], + [ 0.0011, 0.0032, 0.0025, ..., -0.0011, 0.0028, 0.0032], + [ 0.0011, 0.0021, -0.0007, ..., -0.0011, 0.0019, -0.0002], + ..., + [ 0.0522, 0.0205, 0.0304, ..., -0.0522, 0.0039, 0.0275], + [-0.0374, 0.0010, -0.0097, ..., 0.0374, 0.0130, 0.0005], + [-0.0413, 0.0023, 0.0041, ..., 0.0425, 0.0112, 0.0056]], + dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.1.weight': tensor([[-1.2256e-01, -1.2061e-01, -1.1865e-01, ..., 8.8501e-04, + 8.4305e-04, 7.9346e-04], + [ 3.2715e-02, 3.1982e-02, 3.1128e-02, ..., -5.4598e-05, + -1.2279e-05, -7.0095e-05], + [ 2.1118e-02, 2.1606e-02, 2.1973e-02, ..., -1.4973e-04, + -7.5340e-05, -1.0347e-04], + ..., + [-1.3489e-02, -1.2573e-02, -1.1719e-02, ..., 7.6771e-05, + -6.1095e-06, 2.4319e-05], + [ 5.4688e-02, 5.3955e-02, 5.3223e-02, ..., -5.4550e-04, + -5.2261e-04, -5.4550e-04], + [-1.1133e-01, -1.0938e-01, -1.0742e-01, ..., 1.0147e-03, + 1.0071e-03, 1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.24.adaln_modulation_mlp.2.weight': tensor([[ 1.1368e-03, 1.4801e-03, -1.2589e-03, ..., -1.1902e-03, + 3.7689e-03, 1.1520e-03], + [-1.5030e-03, -7.5150e-04, 7.3624e-04, ..., 8.8882e-04, + -1.0157e-04, -1.3428e-03], + [-8.4305e-04, 9.1553e-04, -2.1648e-04, ..., -1.8120e-04, + 2.8076e-03, -1.6098e-03], + ..., + [ 1.0449e-01, 5.0781e-02, -6.8359e-02, ..., -6.0059e-02, + -7.5195e-02, 1.0986e-01], + [ 8.5449e-02, 3.5889e-02, -5.1514e-02, ..., -4.4189e-02, + -1.1035e-01, 9.3750e-02], + [-9.8633e-02, -6.2500e-02, 7.7637e-02, ..., 7.0801e-02, + 7.6660e-02, -1.0107e-01]], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_proj.weight': tensor([[ 0.0056, -0.0036, -0.0068, ..., 0.0032, 0.0137, 0.0095], + [-0.0115, -0.0036, -0.0145, ..., 0.0027, 0.0031, -0.0043], + [-0.0305, -0.0032, -0.0071, ..., -0.0212, 0.0020, 0.0056], + ..., + [-0.0121, -0.0219, 0.0026, ..., -0.0053, 0.0015, -0.0010], + [-0.0019, -0.0095, -0.0165, ..., -0.0125, 0.0052, 0.0014], + [-0.0304, -0.0007, -0.0085, ..., -0.0244, 0.0405, 0.0069]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm.weight': tensor([0.6992, 0.8398, 1.3594, 1.0078, 0.7109, 0.4629, 0.9141, 1.4375, 1.2031, + 1.2188, 1.1172, 1.2656, 1.1406, 1.4297, 1.3047, 1.2344, 1.2578, 1.3125, + 1.3203, 1.2422, 1.3047, 1.0938, 1.4062, 0.4902, 1.7266, 0.6133, 1.1797, + 1.8203, 1.2812, 1.2344, 1.4062, 1.4453, 1.2969, 1.3672, 1.3359, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2734, 1.0703, 0.5898, 1.1406, + 0.4082, 1.7656, 0.7266, 1.5547, 0.9141, 1.0312, 1.3594, 0.7305, 1.3594, + 1.3984, 1.3203, 1.2891, 1.3828, 1.2188, 1.3438, 1.3594, 1.2656, 1.3516, + 1.2578, 1.4141, 1.3438, 0.9141, 0.9180, 0.9375, 0.6758, 1.0625, 1.1484, + 1.1953, 0.1021, 1.3594, 1.3125, 1.2969, 1.2891, 1.1484, 1.3203, 1.1484, + 1.1641, 1.3359, 1.2188, 1.2266, 1.0625, 0.8789, 0.3770, 0.9805, 0.6875, + 1.1797, 1.3516, 1.5078, 1.3438, 1.2812, 1.3359, 1.3672, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.9688, + 1.3203, 0.4805, 1.0859, 0.5391, 1.2031, 1.1875, 1.2734, 1.0859, 1.4219, + 1.4219, 1.2812, 1.2656, 1.3672, 1.2422, 1.3516, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.k_proj.weight': tensor([[ 0.0077, 0.0001, 0.0036, ..., 0.0018, -0.0052, 0.0061], + [ 0.0067, 0.0105, -0.0016, ..., -0.0036, -0.0087, 0.0022], + [-0.0233, 0.0012, 0.0010, ..., -0.0031, 0.0139, 0.0092], + ..., + [-0.0337, -0.0134, -0.0193, ..., -0.0165, 0.0179, -0.0041], + [ 0.0075, -0.0024, -0.0043, ..., -0.0034, 0.0376, 0.0021], + [-0.0136, -0.0107, -0.0050, ..., -0.0063, -0.0042, 0.0020]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm.weight': tensor([1.0234, 1.0859, 1.3281, 0.8203, 0.8555, 0.5430, 0.8906, 1.3906, 1.1797, + 1.2188, 1.1172, 1.2422, 1.1484, 1.4141, 1.3047, 1.2344, 1.2500, 1.3125, + 1.3203, 1.2344, 1.3047, 1.1016, 1.4531, 0.4336, 1.5000, 0.6602, 1.2422, + 1.7344, 1.3516, 1.3359, 1.3906, 1.4297, 1.2891, 1.3672, 1.3281, 1.1406, + 1.3594, 0.1992, 0.6758, 1.2188, 1.3438, 1.2812, 1.0703, 0.9258, 1.1953, + 0.3984, 1.5781, 0.7695, 1.3438, 0.9453, 0.9414, 1.2891, 0.7578, 1.3750, + 1.3984, 1.3047, 1.2891, 1.3828, 1.2266, 1.3438, 1.3594, 1.2578, 1.3672, + 1.2578, 1.3438, 1.2422, 0.9258, 0.9453, 0.8555, 0.6953, 1.0078, 1.2031, + 1.3281, 0.1133, 1.3516, 1.3047, 1.2969, 1.2969, 1.1484, 1.3125, 1.1562, + 1.1641, 1.3359, 1.2188, 1.2344, 1.0625, 0.6992, 0.4785, 1.2734, 0.6445, + 0.9531, 1.3828, 1.4688, 1.3125, 1.2891, 1.3516, 1.3750, 1.3125, 1.2656, + 1.3203, 1.2422, 1.3281, 1.2969, 1.2500, 1.3125, 1.2344, 1.2188, 1.6250, + 1.2422, 0.4844, 1.2500, 0.5000, 1.1953, 1.1797, 1.2891, 1.2656, 1.3594, + 1.4141, 1.2969, 1.2656, 1.3594, 1.2422, 1.3438, 1.3516, 1.2188, 1.3203, + 1.2891, 0.9961], dtype=torch.bfloat16), 'net.blocks.25.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.self_attn.v_proj.weight': tensor([[ 0.0008, -0.0084, 0.0038, ..., 0.0171, -0.0102, 0.0106], + [ 0.0087, 0.0018, -0.0056, ..., 0.0050, 0.0264, 0.0020], + [ 0.0107, 0.0015, -0.0107, ..., -0.0056, 0.0287, 0.0226], + ..., + [ 0.0008, -0.0211, -0.0175, ..., -0.0205, 0.0038, -0.0270], + [-0.0073, -0.0327, 0.0025, ..., -0.0113, 0.0177, 0.0306], + [ 0.0054, 0.0052, -0.0111, ..., -0.0134, -0.0003, 0.0317]], + dtype=torch.bfloat16), 'net.blocks.25.self_attn.output_proj.weight': tensor([[-0.0082, -0.0002, -0.0131, ..., 0.0250, 0.0237, 0.0042], + [ 0.0216, 0.0072, 0.0120, ..., 0.0080, 0.0225, -0.0051], + [ 0.0009, 0.0265, 0.0066, ..., -0.0143, 0.0066, -0.0024], + ..., + [ 0.0110, 0.0077, -0.0039, ..., 0.0095, 0.0189, -0.0117], + [ 0.0053, 0.0276, 0.0074, ..., -0.0038, -0.0221, 0.0157], + [ 0.0019, 0.0045, -0.0082, ..., -0.0019, -0.0078, -0.0071]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_proj.weight': tensor([[-0.0109, 0.0027, 0.0015, ..., -0.0062, -0.0045, 0.0087], + [ 0.0003, 0.0223, -0.0280, ..., -0.0128, 0.0006, -0.0080], + [ 0.0143, -0.0102, 0.0098, ..., -0.0089, 0.0058, -0.0128], + ..., + [-0.0256, -0.0214, 0.0223, ..., 0.0284, 0.0135, -0.0245], + [ 0.0090, -0.0031, 0.0085, ..., -0.0119, 0.0199, -0.0042], + [-0.0226, -0.0157, -0.0209, ..., 0.0099, 0.0165, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm.weight': tensor([ 9.0027e-04, 2.7161e-03, -1.8533e-07, 5.0659e-03, 2.4796e-04, + -1.3097e-10, 2.1515e-03, -1.7062e-06, 1.3785e-12, -1.2398e-08, + 7.3051e-09, 2.5977e-01, 1.0986e-02, 1.4099e-02, 1.6129e-12, + 1.7189e-10, 9.4995e-07, -1.4067e-05, 2.9297e-03, 7.9632e-05, + 2.3651e-03, -6.4233e-12, 1.3965e-01, 7.5531e-04, 4.0894e-03, + -1.1474e-06, -4.3074e-08, 7.9155e-05, -5.1022e-05, 7.1106e-03, + 3.8605e-03, -3.2014e-09, 8.1956e-07, 4.7302e-03, 9.6858e-08, + 3.1891e-03, 4.9210e-04, -5.7199e-13, 5.6458e-03, 3.9062e-01, + 2.1815e-05, 2.7940e-09, 3.9290e-09, 4.2915e-06, -3.2663e-05, + 3.2806e-03, 3.8147e-05, 7.0140e-09, -2.3693e-06, 1.5640e-03, + 1.9791e-09, 3.6240e-05, -2.5332e-06, -8.2655e-09, 5.4932e-03, + 6.5613e-03, -2.5466e-09, 4.5166e-03, 3.9648e-01, 1.3590e-05, + -1.6809e-05, 9.7046e-03, -2.0464e-12, 6.0754e-10, 5.6152e-03, + 4.2725e-03, 3.3528e-08, 5.6458e-04, 2.8381e-03, 1.0254e-02, + -7.4506e-06, 3.1250e-01, -6.0722e-07, 8.8215e-06, -1.0536e-08, + 4.1723e-05, -7.7724e-05, 9.7275e-05, 1.5949e-08, -4.8523e-03, + -1.8239e-05, 2.8229e-04, -1.8190e-10, 3.5286e-05, 2.3499e-03, + 1.0834e-03, -4.7302e-04, 2.1057e-03, 3.1662e-04, 1.3897e-09, + 2.8667e-09, 4.0588e-03, -3.6955e-05, 1.3256e-04, -1.5483e-08, + 2.7649e-10, -1.2577e-05, 6.7444e-03, 1.0550e-09, 6.1035e-05, + 1.6403e-03, 2.5330e-03, 3.6316e-03, -3.1432e-08, 2.3041e-03, + 2.0555e-10, 1.6022e-03, 2.2278e-03, -1.0824e-04, 1.7456e-02, + -8.1491e-09, -2.8968e-05, -8.2850e-06, -2.6375e-06, -2.6659e-08, + 3.2043e-03, 1.9264e-04, 4.0820e-01, -1.4424e-05, 7.7820e-04, + -6.4392e-10, 8.0399e-10, 3.4180e-03, 1.2398e-05, -1.2398e-04, + 6.4087e-03, -3.4343e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_proj.weight': tensor([[ 5.2185e-03, 4.5776e-03, -7.3547e-03, ..., 5.8899e-03, + 5.6458e-03, -8.2970e-05], + [-8.4229e-03, 2.8229e-03, -3.8910e-03, ..., 1.9989e-03, + 3.9978e-03, -9.8419e-04], + [-8.1787e-03, -2.0294e-03, 6.3782e-03, ..., -2.9755e-03, + -7.7515e-03, 7.5073e-03], + ..., + [ 8.9111e-03, -6.7902e-04, -8.4839e-03, ..., -8.9111e-03, + 3.3875e-03, -8.8501e-03], + [-7.7438e-04, 1.3275e-03, 2.6550e-03, ..., 1.2665e-03, + -3.0365e-03, 2.3651e-03], + [ 1.3580e-03, 3.4332e-03, 5.5847e-03, ..., 9.6436e-03, + 2.2888e-03, -4.2419e-03]], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm.weight': tensor([ 8.9645e-04, 2.5177e-03, 1.9372e-07, 5.0659e-03, 1.3924e-04, + -8.8676e-11, 2.1057e-03, 1.7434e-06, -1.4211e-12, -4.5111e-09, + 7.3924e-09, 2.5781e-01, 1.0986e-02, 1.4526e-02, 1.5916e-12, + -3.3288e-10, 9.9838e-07, -1.4663e-05, 2.9907e-03, 5.3883e-05, + 2.5024e-03, 1.2108e-11, 1.3965e-01, 7.4005e-04, 4.1199e-03, + -1.1474e-06, 4.2841e-08, 7.8201e-05, -5.0068e-05, 7.1106e-03, + 3.8605e-03, 3.3906e-09, 8.6799e-07, 4.8523e-03, -9.6392e-08, + 3.1738e-03, 3.7575e-04, 1.9185e-12, 5.5542e-03, 3.8867e-01, + 2.1458e-05, 2.7794e-09, 3.0414e-09, 4.7088e-06, -2.9445e-05, + 3.2959e-03, 9.3937e-05, 1.0186e-08, 1.3858e-06, 1.4954e-03, + -1.8335e-09, 3.5524e-05, 5.1856e-06, -8.6147e-09, 5.5237e-03, + 6.0120e-03, -1.2733e-09, 4.5776e-03, 3.9453e-01, 1.3530e-05, + -1.8716e-05, 9.3994e-03, -2.0179e-12, -2.9286e-10, 5.6458e-03, + 4.2725e-03, -1.0841e-09, 6.1417e-04, 2.8687e-03, 1.0193e-02, + -7.7486e-06, 3.1055e-01, -5.9977e-07, 8.9407e-06, -9.4296e-09, + 1.3530e-05, -7.8201e-05, 9.6798e-05, 1.2631e-08, 4.7913e-03, + -1.7881e-05, -2.8038e-04, 4.5475e-10, 3.3855e-05, 2.0142e-03, + 1.0071e-03, -4.7112e-04, -5.4169e-04, 3.2425e-04, 1.0914e-09, + 2.8376e-09, 4.4250e-03, -3.6240e-05, 1.3351e-04, 7.1246e-08, + 2.8012e-10, -1.2577e-05, 6.6833e-03, -4.6566e-10, 6.1035e-05, + 1.6098e-03, 2.3804e-03, 3.6011e-03, 2.3865e-08, 2.4719e-03, + 2.3192e-10, 1.5869e-03, 2.1973e-03, -1.0824e-04, 1.7456e-02, + 8.4401e-09, -2.8968e-05, -8.4639e-06, -2.7865e-06, 1.6065e-08, + 3.2043e-03, 2.0313e-04, 4.0820e-01, -1.4782e-05, 1.1301e-04, + -7.0213e-10, 8.2946e-10, 3.3875e-03, 1.2398e-05, -1.2779e-04, + 5.1880e-03, 3.5070e-09, 3.5248e-03], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.v_proj.weight': tensor([[-0.0250, -0.0062, -0.0359, ..., -0.0099, -0.0219, -0.0242], + [ 0.0233, 0.0043, 0.0197, ..., 0.0055, 0.0047, 0.0515], + [ 0.0283, 0.0047, 0.0325, ..., 0.0073, 0.0240, 0.0208], + ..., + [-0.0181, 0.0008, -0.0532, ..., -0.0047, -0.0035, -0.0077], + [-0.0134, 0.0038, 0.0192, ..., -0.0026, -0.0040, 0.0040], + [-0.0262, -0.0065, -0.0398, ..., -0.0062, -0.0245, -0.0250]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0053, -0.0088, ..., 0.0045, 0.0048, 0.0016], + [ 0.0108, -0.0116, -0.0189, ..., -0.0010, 0.0038, -0.0092], + [-0.0039, 0.0077, 0.0015, ..., 0.0070, -0.0008, -0.0198], + ..., + [-0.0110, 0.0132, 0.0315, ..., 0.0078, 0.0139, 0.0192], + [-0.0206, 0.0177, 0.0010, ..., -0.0063, 0.0019, -0.0229], + [ 0.0221, -0.0173, -0.0243, ..., 0.0049, -0.0020, -0.0010]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img.weight': tensor([[-0.0264, -0.0371, -0.0248, ..., -0.0471, -0.0393, -0.0139], + [-0.0063, 0.0081, -0.0077, ..., -0.0588, 0.0187, -0.0068], + [-0.0102, 0.0199, -0.0077, ..., 0.0080, -0.0332, -0.0026], + ..., + [-0.0051, -0.0464, 0.0277, ..., -0.0143, 0.0198, 0.0071], + [ 0.0162, -0.0053, 0.0082, ..., -0.0105, -0.0038, 0.0297], + [-0.0269, 0.0129, 0.0164, ..., 0.0086, 0.0176, -0.0042]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.v_img.weight': tensor([[ 0.0208, -0.0300, 0.0033, ..., -0.0505, -0.0332, 0.0153], + [ 0.0208, 0.0366, 0.0028, ..., 0.0142, 0.0128, 0.0439], + [-0.0352, 0.0042, -0.0201, ..., 0.0330, 0.0309, -0.0049], + ..., + [-0.0033, 0.0123, 0.0153, ..., -0.0034, -0.0038, -0.0325], + [ 0.0050, 0.0537, 0.0165, ..., -0.0142, 0.0039, 0.0175], + [-0.0244, 0.0036, -0.0300, ..., 0.0051, 0.0334, -0.0203]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img.weight': tensor([[-0.0231, -0.0046, -0.0019, ..., -0.0109, -0.0130, 0.0359], + [-0.0302, 0.0261, 0.0029, ..., 0.0310, 0.0088, -0.0023], + [ 0.0033, 0.0128, 0.0217, ..., 0.0439, -0.0186, 0.0143], + ..., + [-0.0088, -0.0109, -0.0237, ..., -0.0104, 0.0157, -0.0037], + [-0.0081, 0.0205, 0.0140, ..., 0.0262, -0.0195, 0.0077], + [-0.0183, -0.0270, -0.0364, ..., -0.0601, 0.0176, 0.0077]], + dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.25.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.25.mlp.layer1.weight': tensor([[ 0.0021, -0.0015, 0.0016, ..., -0.0139, 0.0096, 0.0140], + [ 0.0048, 0.0201, 0.0038, ..., -0.0214, -0.0129, -0.0042], + [-0.0047, -0.0226, -0.0090, ..., 0.0420, 0.0188, 0.0154], + ..., + [ 0.0159, -0.0121, 0.0029, ..., -0.0046, 0.0036, 0.0220], + [-0.0067, 0.0049, 0.0104, ..., -0.0020, -0.0030, 0.0045], + [-0.0339, 0.0082, 0.0067, ..., 0.0145, -0.0145, -0.0087]], + dtype=torch.bfloat16), 'net.blocks.25.mlp.layer2.weight': tensor([[-0.0036, 0.0062, 0.0016, ..., 0.0188, -0.0003, -0.0221], + [ 0.0093, -0.0090, -0.0004, ..., -0.0089, -0.0051, -0.0052], + [-0.0137, 0.0093, 0.0115, ..., 0.0212, 0.0132, -0.0090], + ..., + [-0.0150, -0.0119, -0.0110, ..., 0.0079, -0.0152, 0.0071], + [-0.0067, -0.0058, 0.0123, ..., 0.0237, -0.0001, -0.0019], + [-0.0082, 0.0003, -0.0125, ..., -0.0291, -0.0109, -0.0113]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.1.weight': tensor([[-0.0737, -0.0723, -0.0708, ..., 0.0008, 0.0007, 0.0007], + [-0.0688, -0.0674, -0.0659, ..., 0.0006, 0.0006, 0.0007], + [ 0.0771, 0.0752, 0.0742, ..., -0.0006, -0.0005, -0.0005], + ..., + [-0.0737, -0.0723, -0.0703, ..., 0.0007, 0.0007, 0.0007], + [-0.0693, -0.0688, -0.0684, ..., -0.0003, -0.0001, -0.0002], + [ 0.0679, 0.0669, 0.0664, ..., 0.0001, 0.0002, 0.0003]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_self_attn.2.weight': tensor([[ 0.0051, 0.0032, -0.0045, ..., 0.0060, 0.0025, -0.0026], + [-0.0020, -0.0018, 0.0017, ..., -0.0016, 0.0004, -0.0003], + [ 0.0022, 0.0023, -0.0023, ..., 0.0028, 0.0009, -0.0009], + ..., + [-0.0193, -0.0019, 0.0139, ..., -0.0242, -0.0374, 0.0364], + [ 0.0168, -0.0013, -0.0080, ..., 0.0244, 0.0311, -0.0320], + [-0.0216, -0.0046, 0.0144, ..., -0.0270, -0.0369, 0.0374]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.1.weight': tensor([[-9.1553e-03, -9.3994e-03, -9.1553e-03, ..., -1.1492e-04, + -1.5736e-04, -1.3065e-04], + [ 1.4526e-02, 1.4160e-02, 1.3428e-02, ..., -7.6294e-05, + -1.3161e-04, -7.1049e-05], + [-7.5684e-02, -7.4219e-02, -7.2754e-02, ..., 9.1791e-06, + 6.8188e-05, 3.1233e-05], + ..., + [-1.5076e-02, -1.3672e-02, -1.4221e-02, ..., -7.1049e-05, + -1.3542e-04, -1.2159e-04], + [-1.8555e-02, -1.7944e-02, -1.7578e-02, ..., -2.2221e-04, + -1.4687e-04, -8.1062e-05], + [ 2.7344e-02, 2.6367e-02, 2.5879e-02, ..., -9.2506e-05, + -4.5776e-05, -6.6757e-05]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_cross_attn.2.weight': tensor([[-0.0034, -0.0080, 0.0107, ..., -0.0034, -0.0034, -0.0084], + [-0.0162, 0.0075, 0.0082, ..., -0.0167, -0.0166, 0.0074], + [ 0.0044, -0.0018, -0.0032, ..., 0.0046, 0.0045, -0.0019], + ..., + [-0.1055, -0.0461, 0.0137, ..., -0.1099, -0.1074, -0.0405], + [-0.0562, -0.0278, 0.0025, ..., -0.0603, -0.0603, -0.0251], + [ 0.0742, 0.0271, 0.0067, ..., 0.0791, 0.0791, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.1.weight': tensor([[-8.3008e-02, -8.1543e-02, -8.0078e-02, ..., 1.3962e-03, + 1.4420e-03, 1.4572e-03], + [ 1.0010e-01, 9.9121e-02, 9.7656e-02, ..., -5.6505e-05, + -7.2002e-05, -1.0777e-04], + [ 8.4473e-02, 8.3496e-02, 8.2520e-02, ..., -6.0320e-05, + -1.1563e-05, -3.6478e-05], + ..., + [-4.0283e-02, -4.0039e-02, -4.0283e-02, ..., -2.9755e-04, + -3.7909e-05, -6.3419e-05], + [ 9.4727e-02, 9.3262e-02, 9.2285e-02, ..., -1.6499e-04, + -6.4850e-05, -1.3924e-04], + [ 8.6914e-02, 8.4473e-02, 8.2520e-02, ..., -1.1902e-03, + -1.2436e-03, -1.2131e-03]], dtype=torch.bfloat16), 'net.blocks.25.adaln_modulation_mlp.2.weight': tensor([[-8.4839e-03, -1.5030e-03, -1.5793e-03, ..., 1.5717e-03, + -1.3657e-03, 6.7749e-03], + [-3.5553e-03, -3.4180e-03, -3.7231e-03, ..., 4.3335e-03, + -3.6926e-03, 2.5787e-03], + [-3.4332e-04, -7.2002e-05, -8.5831e-05, ..., -6.1035e-05, + -1.6022e-04, -3.5095e-04], + ..., + [ 5.8838e-02, -1.0498e-01, -1.0303e-01, ..., 9.7656e-02, + -1.0449e-01, -7.9590e-02], + [ 8.4473e-02, -8.8867e-02, -8.5938e-02, ..., 8.2031e-02, + -8.5938e-02, -9.4727e-02], + [-9.4238e-02, 1.0498e-01, 1.0303e-01, ..., -9.5215e-02, + 1.0205e-01, 1.0254e-01]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_proj.weight': tensor([[ 0.0003, -0.0081, 0.0020, ..., -0.0085, 0.0065, 0.0107], + [ 0.0035, 0.0002, 0.0039, ..., 0.0093, -0.0019, -0.0031], + [-0.0011, -0.0090, -0.0006, ..., -0.0055, -0.0089, 0.0049], + ..., + [-0.0060, 0.0059, -0.0043, ..., 0.0084, -0.0221, 0.0096], + [ 0.0038, 0.0079, 0.0159, ..., -0.0067, 0.0041, -0.0091], + [ 0.0078, -0.0100, 0.0157, ..., 0.0060, -0.0159, -0.0206]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm.weight': tensor([ 1.8203, 1.2578, 1.4844, 0.9375, 1.0547, 1.5938, 1.2812, 1.1328, + 1.0078, 1.4375, 1.2422, 1.3281, 1.3516, 1.2422, 1.1562, 1.5078, + 1.4375, 1.2578, 1.3203, 1.1797, 1.1484, 1.2031, 1.5703, 1.7031, + 1.4375, 0.0894, 1.0078, 1.5859, 1.2500, 1.2969, 1.5469, 1.3438, + 1.4297, 1.4297, 1.4844, 1.3984, 1.4453, 1.3516, 1.4609, 1.2656, + 1.2578, 1.2500, 1.3125, 0.8672, 1.8359, 1.7109, 1.1797, -0.0187, + 1.3359, 1.2969, 1.5000, 1.2969, 1.1797, 1.4609, 1.2344, 1.4766, + 1.2656, 1.3594, 1.4375, 1.4297, 1.2500, 1.2188, 1.2500, 1.2656, + 1.1250, 1.7344, 1.4531, 1.6328, 1.4375, 1.1875, 1.4297, 1.1484, + 1.5859, 1.4141, 1.2734, 1.6016, 1.3828, 1.3828, 1.3047, 1.3906, + 1.2812, 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.8672, 1.4062, + 1.3203, 0.2676, 1.0859, 1.7031, 1.5469, 1.4844, 1.1797, 1.2422, + 1.2578, 1.3516, 1.3047, 1.1328, 1.2422, 1.5547, 1.2891, 1.1953, + 1.3672, 1.2266, 1.1172, 1.8594, 1.5938, 1.1719, 1.4062, 0.8438, + 1.7344, 1.3984, 1.2969, 1.4297, 1.4375, 1.2578, 1.3750, 1.2109, + 1.2891, 1.5156, 1.3828, 1.4453, 1.3359, 1.3047, 1.2656, 1.2031], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.k_proj.weight': tensor([[ 0.0203, -0.0011, -0.0027, ..., 0.0003, 0.0129, -0.0013], + [ 0.0096, 0.0055, -0.0156, ..., -0.0004, -0.0170, -0.0071], + [ 0.0256, 0.0056, 0.0063, ..., 0.0020, -0.0038, 0.0099], + ..., + [-0.0006, 0.0029, -0.0128, ..., -0.0044, -0.0151, 0.0143], + [ 0.0154, -0.0030, 0.0040, ..., 0.0108, 0.0120, 0.0136], + [ 0.0036, 0.0051, -0.0161, ..., 0.0039, 0.0317, 0.0043]], + dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm.weight': tensor([1.0391, 1.4609, 1.4688, 1.1172, 1.2656, 1.4766, 1.3438, 1.1172, 1.2109, + 1.4297, 1.2344, 1.3906, 1.3516, 1.2578, 1.1562, 1.5078, 1.4375, 1.2656, + 1.3281, 1.1797, 1.1484, 1.2031, 1.6719, 1.7266, 1.3516, 0.1045, 0.8945, + 1.5781, 1.3125, 1.3594, 1.5391, 1.3516, 1.4219, 1.4297, 1.4844, 1.3984, + 1.4531, 1.3516, 1.4609, 1.2656, 1.2656, 1.2500, 1.3125, 1.0469, 1.8203, + 1.6484, 1.1641, 0.0156, 1.3438, 1.1953, 1.4609, 1.3125, 1.2422, 1.4609, + 1.2500, 1.4766, 1.2656, 1.3672, 1.4375, 1.4219, 1.2500, 1.2188, 1.2578, + 1.2656, 1.4766, 1.1797, 1.2656, 1.5781, 1.1719, 1.3906, 1.4062, 1.1562, + 1.5703, 1.4219, 1.2734, 1.5703, 1.3828, 1.3828, 1.3047, 1.3906, 1.2656, + 1.3359, 1.2500, 1.3125, 1.2031, 1.1016, 1.4141, 1.4141, 1.3594, 0.2578, + 1.0859, 1.7109, 1.5469, 1.4688, 1.2422, 1.2500, 1.2578, 1.3516, 1.2891, + 1.1328, 1.2656, 1.5547, 1.2891, 1.1953, 1.3672, 1.2344, 1.1172, 1.8281, + 1.5391, 1.2422, 1.3828, 0.8398, 1.6875, 1.3828, 1.3438, 1.4062, 1.4297, + 1.2969, 1.3672, 1.2266, 1.2891, 1.5156, 1.3828, 1.4609, 1.3359, 1.3047, + 1.2656, 1.2031], dtype=torch.bfloat16), 'net.blocks.26.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.self_attn.v_proj.weight': tensor([[-4.6692e-03, -1.6403e-03, -5.2734e-02, ..., -1.3611e-02, + -1.3794e-02, -1.2024e-02], + [ 1.1414e-02, -3.0518e-02, -1.2146e-02, ..., 2.7008e-03, + -2.3926e-02, -2.1973e-02], + [-3.5095e-03, -2.0386e-02, -1.3184e-02, ..., -7.1716e-03, + 2.0874e-02, -6.0120e-03], + ..., + [-5.0049e-03, -2.5635e-02, 3.8757e-03, ..., 1.0742e-02, + -1.0529e-03, 8.6670e-03], + [-1.3649e-05, -7.1411e-03, 9.4604e-03, ..., 1.5926e-04, + 3.9062e-02, -2.0996e-02], + [ 1.2756e-02, -4.8828e-03, 2.3315e-02, ..., 2.6489e-02, + 6.4697e-03, 7.6599e-03]], dtype=torch.bfloat16), 'net.blocks.26.self_attn.output_proj.weight': tensor([[ 1.1780e-02, 5.6458e-03, -1.2024e-02, ..., -1.1719e-02, + -4.1504e-03, -3.9368e-03], + [ 1.8799e-02, 1.3123e-02, -5.6076e-04, ..., 7.5684e-03, + 6.8970e-03, 2.1484e-02], + [-1.9531e-02, -9.9487e-03, -8.4229e-03, ..., 1.1108e-02, + -2.4902e-02, 7.8125e-03], + ..., + [ 1.0193e-02, 7.7248e-05, 1.4832e-02, ..., 6.9580e-03, + 7.9956e-03, 1.1841e-02], + [-1.4221e-02, -1.4954e-02, -8.3618e-03, ..., 7.1411e-03, + 4.2419e-03, 1.0437e-02], + [-2.6894e-04, -3.3875e-03, -3.8300e-03, ..., 8.3008e-03, + 5.1575e-03, 1.8799e-02]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_proj.weight': tensor([[-8.3618e-03, -4.3945e-03, -8.3008e-03, ..., 5.0659e-03, + -1.9287e-02, 2.3346e-03], + [-1.2756e-02, 2.1729e-02, -4.3030e-03, ..., -2.6367e-02, + 2.2339e-02, 6.8665e-03], + [-2.4170e-02, -1.7090e-02, -2.3804e-02, ..., -1.1597e-02, + -8.3008e-03, -4.5471e-03], + ..., + [ 3.9978e-03, 1.1536e-02, 1.2695e-02, ..., -1.1292e-02, + 1.4343e-02, 3.7551e-06], + [ 2.8839e-03, 1.5564e-02, -2.4567e-03, ..., -5.0964e-03, + 2.4719e-03, 1.5411e-03], + [ 3.2654e-03, -7.8125e-03, 3.0884e-02, ..., -8.9722e-03, + -2.2217e-02, 7.9346e-03]], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm.weight': tensor([ 2.9755e-03, -7.3426e-34, -3.1537e-33, 1.4324e-33, -6.2593e-33, + 7.7519e-33, 8.7871e-34, 1.2338e-33, -1.4565e-33, 2.2705e-02, + 4.5645e-32, 2.6331e-34, 3.9963e-33, 1.1755e-35, 1.0437e-02, + -5.8019e-33, 1.8597e-05, 2.8438e-34, 9.2084e-34, -9.3287e-34, + -1.8417e-33, -4.3093e-33, -1.8537e-33, -2.5519e-33, 3.4359e-31, + 2.6971e-06, 3.3742e-31, -1.6178e-32, -2.5759e-33, -4.1612e-29, + 3.7384e-03, 2.2389e-33, 2.3365e-05, 2.5391e-02, -8.4803e-30, + -4.4838e-34, -4.0445e-33, 4.1797e-01, -4.5557e-29, -1.7153e-34, + -8.6185e-33, 1.8358e-05, -1.0689e-32, -2.4194e-33, 2.4805e-01, + -4.4768e-29, -1.6479e-03, 4.0430e-01, 1.2779e-04, -1.5648e-33, + 3.4757e-34, 2.4194e-33, -4.4763e-35, 3.6133e-01, 1.5777e-29, + -4.8148e-33, 4.4250e-04, 7.1228e-06, -3.0093e-33, -1.8657e-33, + 1.3963e-32, -7.8843e-34, -1.8082e-03, -9.1508e-29, 1.7524e-05, + 4.1602e-01, 2.4676e-33, -4.6704e-33, 1.7213e-33, -3.4185e-33, + -1.9722e-31, 5.8741e-33, 6.5903e-34, 1.5259e-05, -2.1763e-32, + -1.5648e-33, 1.6756e-32, -4.6042e-34, 7.2021e-03, -2.1366e-34, + 6.7139e-03, -2.6253e-27, 9.3132e-10, -1.3181e-33, -7.1621e-34, + 6.6926e-33, -3.1597e-34, 2.3472e-34, 5.4167e-33, 4.9770e-06, + -3.4907e-33, 1.8921e-02, -1.2579e-33, 4.7245e-34, -1.3199e-03, + 4.0984e-31, 2.0996e-01, 4.4778e-33, -2.3593e-33, -4.1227e-34, + -4.3093e-33, 1.7822e-02, -3.1778e-33, -2.8046e-33, -1.3090e-34, + 1.7395e-03, 1.0292e-33, 2.7565e-33, -4.2431e-34, 1.4033e-13, + 2.0742e-05, -6.2894e-34, 1.1444e-05, 9.5156e-30, -1.1360e-26, + 1.0111e-33, 1.4267e-03, -1.1255e-33, -5.6453e-30, -2.6361e-33, + 1.2157e-33, 9.3579e-06, -5.5852e-33, -7.6074e-33, 1.3617e-34, + 7.4463e-03, -9.6560e-06, -6.3195e-34], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_proj.weight': tensor([[-0.0095, -0.0066, -0.0004, ..., 0.0011, 0.0002, -0.0009], + [-0.0018, -0.0039, -0.0006, ..., -0.0034, 0.0030, -0.0026], + [-0.0075, 0.0042, 0.0041, ..., -0.0002, 0.0054, -0.0022], + ..., + [-0.0070, 0.0038, 0.0016, ..., 0.0025, 0.0047, -0.0045], + [ 0.0012, 0.0027, 0.0015, ..., -0.0022, -0.0058, 0.0095], + [-0.0010, -0.0042, -0.0006, ..., 0.0019, -0.0059, -0.0024]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm.weight': tensor([ 2.7924e-03, 1.6732e-33, 2.8648e-33, 2.7324e-33, -4.5259e-33, + 5.8982e-33, 4.8195e-36, 1.5723e-34, 5.3264e-34, 2.2705e-02, + 5.1759e-33, -1.6250e-33, -5.6093e-33, -7.3727e-35, 9.8267e-03, + -6.5482e-33, 1.8716e-05, -5.6875e-34, -1.9139e-33, 1.9259e-34, + -3.0815e-33, 1.9620e-33, -3.5870e-33, 2.9611e-33, 9.0519e-32, + 3.1292e-06, -7.8482e-33, -1.9452e-32, 4.5259e-33, -4.0232e-29, + 2.1118e-02, 3.2500e-33, 2.4915e-05, 2.3071e-02, -2.2926e-30, + 1.1375e-33, -9.9185e-33, 4.1406e-01, -4.3782e-29, -1.2519e-33, + 1.0207e-32, 1.8358e-05, 4.7185e-33, -1.2579e-33, 2.4707e-01, + 2.2581e-29, 5.1270e-03, 3.9844e-01, 1.2875e-04, -2.1667e-33, + -1.6972e-33, 5.8019e-33, -3.7766e-34, 3.5938e-01, -1.5087e-29, + 4.3093e-33, 4.8447e-04, 7.9274e-06, -7.0778e-33, 4.1528e-34, + -8.4741e-32, -7.8241e-34, 1.9409e-02, -9.5058e-29, 1.8835e-05, + 4.1602e-01, -5.0315e-33, 3.3704e-33, 5.3791e-35, 1.7153e-34, + 2.6747e-30, -1.3963e-33, -2.1907e-33, 1.5199e-05, 8.7871e-34, + -7.5111e-33, 8.8111e-33, -4.8389e-33, 1.9653e-02, 8.1250e-34, + 6.9580e-03, -2.6506e-27, 4.6157e-11, -1.5167e-33, 1.7454e-33, + -6.5301e-34, 4.4778e-33, -6.8611e-34, -7.7037e-33, 4.4107e-06, + -5.1278e-33, 1.8311e-02, 4.4537e-34, -8.1250e-34, 1.0498e-02, + 1.3404e-31, 2.0898e-01, -3.2500e-33, -1.1435e-33, -8.9074e-34, + -3.2019e-33, 1.7822e-02, 1.6370e-33, 3.3704e-33, -1.7093e-33, + 1.7471e-03, 3.9963e-33, 3.0454e-33, -2.0613e-34, 4.6629e-14, + 1.9073e-05, 9.9306e-34, 1.3649e-05, -6.9518e-30, 1.0400e-26, + 5.3791e-35, 1.6174e-03, 4.0685e-33, -2.9120e-31, 5.1037e-33, + 5.4167e-33, 9.7752e-06, 6.5963e-33, -2.3954e-33, -4.5259e-33, + 7.4463e-03, -9.7752e-06, -1.0051e-33], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.v_proj.weight': tensor([[ 0.0198, 0.0079, 0.0287, ..., 0.0089, 0.0120, 0.0342], + [-0.0030, 0.0052, 0.0139, ..., -0.0005, -0.0046, 0.0166], + [-0.0245, -0.0063, -0.0315, ..., -0.0073, -0.0236, -0.0304], + ..., + [-0.0164, -0.0089, -0.0243, ..., -0.0071, -0.0177, -0.0300], + [ 0.0159, 0.0099, 0.0208, ..., 0.0087, 0.0197, 0.0291], + [-0.0103, -0.0104, -0.0198, ..., -0.0074, -0.0131, -0.0270]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.output_proj.weight': tensor([[-0.0012, -0.0211, -0.0009, ..., -0.0272, 0.0275, -0.0260], + [-0.0032, 0.0030, -0.0036, ..., 0.0178, -0.0249, 0.0248], + [-0.0206, -0.0154, 0.0200, ..., -0.0063, 0.0078, -0.0041], + ..., + [ 0.0100, -0.0041, -0.0082, ..., -0.0130, 0.0095, -0.0009], + [ 0.0076, 0.0081, -0.0071, ..., 0.0111, -0.0104, 0.0010], + [-0.0292, 0.0030, 0.0135, ..., 0.0042, -0.0107, 0.0129]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img.weight': tensor([[-0.0258, 0.0105, 0.0040, ..., -0.0052, 0.0219, -0.0271], + [-0.0352, 0.0041, 0.0221, ..., -0.0018, 0.0034, -0.0093], + [-0.0009, 0.0019, 0.0398, ..., -0.0304, -0.0029, 0.0576], + ..., + [-0.0315, -0.0082, 0.0253, ..., -0.0097, -0.0287, 0.0153], + [-0.0210, 0.0444, -0.0146, ..., -0.0476, -0.0154, -0.0019], + [ 0.0352, -0.0576, 0.0050, ..., 0.0194, 0.0098, 0.0184]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.v_img.weight': tensor([[ 0.0042, 0.0065, 0.0228, ..., -0.0322, 0.0315, -0.0012], + [ 0.0325, -0.0219, 0.0085, ..., 0.0269, 0.0050, 0.0291], + [ 0.0135, -0.0014, -0.0339, ..., 0.0074, -0.0135, -0.0097], + ..., + [-0.0173, 0.0212, -0.0101, ..., -0.0444, 0.0393, -0.0024], + [-0.0166, -0.0043, -0.0430, ..., -0.0167, -0.0234, -0.0210], + [ 0.0250, 0.0065, -0.0039, ..., 0.0240, 0.0037, 0.0012]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img.weight': tensor([[ 0.0337, -0.0079, 0.0410, ..., -0.0039, 0.0096, -0.0020], + [-0.0033, 0.0211, 0.0422, ..., 0.0120, -0.0088, 0.0114], + [ 0.0244, -0.0349, 0.0393, ..., 0.0056, -0.0153, 0.0058], + ..., + [-0.0051, 0.0236, 0.0014, ..., 0.0126, 0.0349, -0.0170], + [ 0.0171, -0.0488, 0.0114, ..., 0.0040, 0.0030, 0.0151], + [-0.0011, -0.0122, 0.0244, ..., -0.0044, -0.0601, 0.0259]], + dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.26.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.26.mlp.layer1.weight': tensor([[-0.0371, -0.0047, -0.0057, ..., 0.0090, -0.0151, 0.0035], + [-0.0210, -0.0036, -0.0004, ..., 0.0210, 0.0002, 0.0003], + [ 0.0082, 0.0065, 0.0067, ..., -0.0103, 0.0125, -0.0049], + ..., + [ 0.0138, -0.0009, 0.0320, ..., 0.0065, 0.0022, -0.0044], + [-0.0092, -0.0094, -0.0049, ..., 0.0140, 0.0145, 0.0145], + [ 0.0081, -0.0248, -0.0046, ..., 0.0415, -0.0078, 0.0157]], + dtype=torch.bfloat16), 'net.blocks.26.mlp.layer2.weight': tensor([[-0.0137, 0.0040, 0.0186, ..., 0.0400, -0.0033, 0.0031], + [ 0.0056, 0.0057, -0.0022, ..., -0.0110, 0.0065, 0.0038], + [-0.0031, -0.0082, 0.0236, ..., 0.0132, 0.0052, -0.0103], + ..., + [-0.0171, 0.0029, -0.0044, ..., 0.0153, 0.0060, -0.0047], + [-0.0053, -0.0031, 0.0110, ..., 0.0039, 0.0115, 0.0020], + [ 0.0093, 0.0033, 0.0034, ..., -0.0032, -0.0128, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.1.weight': tensor([[ 6.5430e-02, 6.4453e-02, 6.3477e-02, ..., 9.7752e-05, + -4.2677e-05, -8.2850e-06], + [-6.8359e-02, -6.7383e-02, -6.5918e-02, ..., 1.7071e-04, + 1.9550e-04, 2.3460e-04], + [ 6.6895e-02, 6.5918e-02, 6.5430e-02, ..., -3.1471e-04, + -3.4142e-04, -3.7003e-04], + ..., + [-7.4707e-02, -7.3242e-02, -7.1289e-02, ..., 8.3923e-04, + 8.3923e-04, 7.7438e-04], + [ 4.2114e-03, 2.6703e-03, -1.1978e-03, ..., -7.2479e-04, + -8.1635e-04, -7.7820e-04], + [-6.3477e-02, -6.2988e-02, -6.2012e-02, ..., 1.6479e-03, + 1.6632e-03, 1.6174e-03]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_self_attn.2.weight': tensor([[-4.5166e-03, 3.8605e-03, -3.7994e-03, ..., 5.4932e-03, + 3.6163e-03, 8.6060e-03], + [-6.4468e-04, -8.8692e-05, -4.2617e-06, ..., -6.7139e-04, + 3.5095e-03, -1.8539e-03], + [ 1.3657e-03, -1.0681e-03, 1.0452e-03, ..., 1.1139e-03, + -1.1902e-03, 4.1809e-03], + ..., + [ 4.8340e-02, -3.9795e-02, 3.8818e-02, ..., -1.8555e-02, + -5.9814e-02, 2.0752e-02], + [-4.1504e-02, 4.1016e-02, -4.0283e-02, ..., 3.8574e-02, + 3.8574e-02, 2.8320e-02], + [ 4.6387e-02, -4.3457e-02, 4.3213e-02, ..., -2.8687e-02, + -4.6631e-02, -9.7656e-04]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.1.weight': tensor([[ 5.8350e-02, 5.6885e-02, 5.5420e-02, ..., -1.1015e-04, + -8.6308e-05, -3.1948e-05], + [-9.7168e-02, -9.4727e-02, -9.2285e-02, ..., 6.6757e-05, + 1.1969e-04, 5.9843e-05], + [ 8.2520e-02, 7.9590e-02, 7.7637e-02, ..., 1.0633e-04, + -6.1989e-05, 1.5646e-06], + ..., + [ 1.9775e-02, 1.9043e-02, 1.8188e-02, ..., -6.0320e-05, + 4.6015e-05, -7.8201e-05], + [-9.8633e-02, -9.5703e-02, -9.4238e-02, ..., 2.6345e-05, + 1.8954e-05, 1.0395e-04], + [ 9.4727e-02, 9.2773e-02, 9.0332e-02, ..., -5.6982e-05, + -8.1062e-05, 4.7386e-06]], dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_cross_attn.2.weight': tensor([[-0.0025, -0.0040, 0.0062, ..., 0.0017, -0.0049, 0.0027], + [-0.0055, -0.0042, 0.0071, ..., 0.0076, -0.0095, 0.0075], + [ 0.0012, -0.0097, 0.0101, ..., 0.0016, -0.0122, 0.0106], + ..., + [-0.0019, 0.0075, 0.0010, ..., 0.0209, 0.0107, -0.0100], + [ 0.0009, 0.0104, -0.0221, ..., -0.0305, 0.0083, -0.0067], + [-0.0013, 0.0070, 0.0012, ..., 0.0195, 0.0093, -0.0093]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.1.weight': tensor([[-0.0334, -0.0322, -0.0308, ..., 0.0007, 0.0008, 0.0009], + [-0.0913, -0.0903, -0.0889, ..., 0.0008, 0.0008, 0.0007], + [ 0.0874, 0.0869, 0.0854, ..., -0.0007, -0.0008, -0.0008], + ..., + [-0.0938, -0.0923, -0.0918, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0359, 0.0349, ..., -0.0007, -0.0008, -0.0007], + [ 0.1079, 0.1060, 0.1040, ..., -0.0016, -0.0016, -0.0017]], + dtype=torch.bfloat16), 'net.blocks.26.adaln_modulation_mlp.2.weight': tensor([[-0.0008, 0.0013, -0.0009, ..., 0.0008, 0.0007, 0.0005], + [ 0.0009, -0.0009, 0.0013, ..., -0.0013, -0.0009, 0.0020], + [ 0.0006, -0.0018, 0.0007, ..., -0.0018, -0.0006, 0.0075], + ..., + [-0.1187, 0.1436, -0.1572, ..., 0.1562, 0.1167, -0.0718], + [-0.1162, 0.1064, -0.1196, ..., 0.1201, 0.1157, -0.0243], + [ 0.0962, -0.1406, 0.1396, ..., -0.1357, -0.0933, 0.0879]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_proj.weight': tensor([[ 0.0040, 0.0007, 0.0054, ..., 0.0006, -0.0054, -0.0029], + [-0.0026, -0.0025, 0.0030, ..., -0.0027, -0.0030, -0.0038], + [-0.0053, -0.0082, 0.0046, ..., -0.0131, -0.0038, 0.0054], + ..., + [ 0.0080, 0.0016, 0.0075, ..., -0.0041, 0.0143, -0.0022], + [ 0.0118, 0.0026, 0.0024, ..., -0.0111, 0.0025, -0.0033], + [-0.0095, 0.0036, 0.0060, ..., 0.0048, -0.0074, -0.0026]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm.weight': tensor([2.2812, 2.0938, 1.3438, 0.6484, 0.8398, 0.0830, 0.4082, 0.3613, 0.0562, + 0.3633, 0.1982, 0.6055, 0.4805, 0.1099, 0.2520, 0.3633, 0.0220, 0.6133, + 0.0596, 0.3691, 1.0938, 0.4141, 1.8672, 1.7266, 0.2090, 0.6680, 1.1953, + 0.0679, 0.5586, 0.2793, 0.6719, 0.3379, 0.4805, 0.5352, 0.0562, 0.6992, + 0.2969, 0.3457, 0.0620, 0.1387, 0.1069, 0.4238, 0.0786, 2.0156, 1.6328, + 1.3438, 1.5781, 0.2412, 1.4922, 1.3594, 1.2969, 0.4922, 0.6016, 0.4785, + 0.3633, 0.2852, 0.1230, 0.3730, 0.0503, 0.3320, 0.0618, 0.0503, 0.0574, + 0.7070, 1.7188, 1.5391, 1.5781, 1.3672, 0.7070, 0.0479, 1.1953, 1.0625, + 0.0649, 1.0703, 0.0972, 0.1118, 0.0591, 0.4883, 0.1406, 0.0374, 0.3730, + 0.1416, 0.3730, 0.4902, 0.0190, 0.4805, 0.6289, 1.3047, 0.1699, 0.7188, + 1.5781, 0.1250, 0.1611, 0.0840, 0.1699, 1.1641, 0.2197, 0.0500, 0.4316, + 0.4531, 0.0806, 0.2676, 0.1104, 0.2949, 0.0466, 0.3906, 0.2500, 1.7500, + 1.7422, 1.3672, 1.5938, 0.1895, 1.7344, 1.1562, 1.2422, 1.2578, 0.2441, + 0.4277, 0.0679, 0.5703, 0.0391, 0.1206, 0.1074, 0.2617, 0.6602, 0.1060, + 0.2656, 0.0275], dtype=torch.bfloat16), 'net.blocks.27.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.k_proj.weight': tensor([[-2.0508e-02, 5.7373e-03, -7.7438e-04, ..., 6.3324e-04, + -1.7471e-03, 1.3367e-02], + [ 4.5471e-03, -3.7384e-03, 2.3651e-03, ..., -4.0894e-03, + -6.1512e-05, -1.9043e-02], + [-2.6245e-03, 2.8076e-03, 1.7548e-03, ..., -1.0071e-02, + 3.2806e-03, -2.1057e-03], + ..., + [-9.3994e-03, -4.0588e-03, 8.7891e-03, ..., -1.0437e-02, + 1.1780e-02, -5.4016e-03], + [-7.6294e-03, -8.5068e-04, 7.7515e-03, ..., -1.6968e-02, + -1.3306e-02, -1.3306e-02], + [ 3.1891e-03, 1.9684e-03, 1.5625e-02, ..., -1.1719e-02, + -1.9455e-03, -1.0071e-03]], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm.weight': tensor([0.8867, 1.6641, 1.6250, 0.6289, 0.8281, 0.0571, 0.0591, 0.0625, 0.0352, + 0.0796, 0.1992, 0.6055, 0.4805, 0.1074, 0.2500, 0.3633, 0.0228, 0.6055, + 0.0588, 0.3711, 1.0938, 0.4141, 1.7500, 1.6172, 0.1992, 0.6953, 1.2109, + 0.1006, 0.5625, 0.2852, 0.6680, 0.2930, 0.4805, 0.5312, 0.0422, 0.6992, + 0.2969, 0.3457, 0.0625, 0.1367, 0.1079, 0.4238, 0.0786, 1.2891, 1.6953, + 1.3438, 1.7500, 0.2422, 1.4766, 1.3359, 1.2891, 0.5039, 0.6016, 0.4805, + 0.3633, 0.2871, 0.1245, 0.3730, 0.0498, 0.3340, 0.0591, 0.0508, 0.0574, + 0.7070, 2.0312, 2.1094, 1.4062, 1.3750, 0.6328, 0.0718, 1.2031, 1.0859, + 0.0708, 1.0781, 0.0957, 0.1104, 0.0645, 0.4883, 0.1387, 0.0391, 0.3711, + 0.1416, 0.3730, 0.4902, 0.0226, 0.4805, 1.9609, 1.6953, 0.1807, 0.5391, + 1.3828, 0.1318, 0.0522, 0.0762, 0.0136, 1.1641, 0.2217, 0.0588, 0.4336, + 0.4453, 0.0776, 0.2676, 0.1104, 0.2949, 0.0471, 0.3906, 0.2500, 1.7969, + 1.6719, 1.4297, 1.5469, 0.2021, 1.7500, 1.1641, 1.2734, 1.2578, 0.0033, + 0.3613, 0.0630, 0.5625, 0.0371, 0.1216, 0.1089, 0.2617, 0.6641, 0.1060, + 0.2637, 0.0267], dtype=torch.bfloat16), 'net.blocks.27.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.self_attn.v_proj.weight': tensor([[-0.0006, 0.0037, 0.0574, ..., -0.0106, -0.0004, 0.0014], + [-0.0067, -0.0052, 0.0017, ..., -0.0151, 0.0109, -0.0092], + [ 0.0044, 0.0087, 0.0029, ..., 0.0039, 0.0063, -0.0179], + ..., + [-0.0091, -0.0036, 0.0007, ..., -0.0134, 0.0074, 0.0055], + [ 0.0045, 0.0046, -0.0026, ..., 0.0050, -0.0063, -0.0067], + [ 0.0112, -0.0021, 0.0059, ..., 0.0110, -0.0023, -0.0004]], + dtype=torch.bfloat16), 'net.blocks.27.self_attn.output_proj.weight': tensor([[-4.5013e-04, -1.6235e-02, 2.1057e-03, ..., 1.2634e-02, + 3.7109e-02, 2.3071e-02], + [ 1.8799e-02, -3.1433e-03, 5.5552e-05, ..., -3.8757e-03, + 1.5320e-02, -4.3640e-03], + [ 7.0801e-03, 3.8910e-03, 9.2163e-03, ..., 2.9297e-02, + -1.7090e-02, -3.0273e-02], + ..., + [ 1.7090e-02, 1.0681e-02, 1.1292e-03, ..., -5.2185e-03, + 5.3406e-03, -1.5991e-02], + [ 1.8066e-02, -2.0996e-02, -1.0803e-02, ..., 5.8289e-03, + -1.4404e-02, -8.8501e-03], + [ 9.0942e-03, 1.2695e-02, -9.9487e-03, ..., -2.4414e-02, + 1.8677e-02, -1.3794e-02]], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_proj.weight': tensor([[ 0.0037, 0.0090, 0.0109, ..., -0.0049, 0.0232, -0.0182], + [-0.0270, -0.0223, 0.0022, ..., -0.0011, -0.0095, 0.0177], + [-0.0131, 0.0017, -0.0025, ..., -0.0042, 0.0060, -0.0062], + ..., + [-0.0187, -0.0039, -0.0040, ..., 0.0053, 0.0082, -0.0019], + [ 0.0170, 0.0094, 0.0090, ..., -0.0058, 0.0280, 0.0173], + [ 0.0175, -0.0155, -0.0099, ..., -0.0036, 0.0078, 0.0066]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm.weight': tensor([ 1.2207e-02, 4.6541e-13, -1.1496e-09, -4.1008e-04, 1.0505e-10, + 8.2397e-03, -3.4535e-23, 4.5967e-04, -1.1511e-12, 3.5553e-03, + -1.7602e-07, 2.2461e-02, 3.3760e-08, 2.0142e-02, -4.4471e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3691e-02, 3.6694e-07, + 2.0862e-06, 7.0801e-03, 6.3281e-01, 2.7148e-01, -2.9569e-08, + -1.1206e-04, 2.4680e-08, 1.0848e-05, 4.0430e-01, -2.9144e-03, + -7.8580e-09, 6.1840e-07, 1.3733e-04, 8.6427e-07, -7.7300e-08, + -1.3486e-06, 1.1475e-02, -4.3154e-05, -8.1855e-11, 7.2718e-06, + 1.1921e-07, -8.9645e-04, 4.7874e-04, -8.8811e-06, 7.0801e-03, + -9.2163e-03, -3.6806e-06, 2.2217e-02, -1.8716e-05, 3.5286e-04, + 8.5449e-03, -2.9989e-07, 2.8198e-02, 6.1646e-03, -3.8708e-09, + 1.2695e-02, 4.2617e-06, 1.3351e-05, 2.9541e-02, 4.6082e-03, + 3.1948e-05, 2.2411e-05, 2.4438e-06, -2.6971e-06, 1.4771e-02, + 2.6978e-02, 8.0466e-06, 7.2632e-03, 1.8001e-05, 1.9550e-05, + 2.3438e-02, 2.4319e-05, 1.1866e-12, -1.0347e-04, 1.1502e-07, + -7.9274e-06, 8.6670e-03, 2.6245e-02, 1.1169e-02, 3.0029e-02, + 6.8054e-03, 1.2517e-06, -9.0804e-09, 8.1491e-08, 2.3193e-02, + 1.3504e-03, 3.2959e-02, 1.1846e-06, 1.2512e-02, 2.5988e-05, + 2.3167e-08, 2.0386e-02, -1.5423e-06, -2.7313e-03, -3.0122e-09, + 4.0527e-02, -1.8835e-05, -3.5652e-09, 3.4925e-09, 1.8539e-03, + 2.8372e-05, 2.3071e-02, -2.9907e-02, -5.5730e-06, 8.3618e-03, + -7.0632e-06, -9.6436e-03, -2.8419e-04, 2.0630e-02, 6.7139e-03, + -1.5571e-09, 1.4249e-07, 7.1335e-04, 9.1553e-03, 3.0365e-03, + 1.2457e-05, 2.1606e-02, 2.0752e-02, 2.6855e-02, 3.9935e-06, + 8.1956e-08, 3.5248e-03, 1.4782e-05, 5.6624e-06, 4.3701e-02, + -4.1580e-04, 6.4611e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_proj.weight': tensor([[ 0.0075, -0.0147, 0.0216, ..., -0.0212, 0.0146, -0.0147], + [-0.0074, -0.0103, 0.0164, ..., -0.0168, 0.0131, 0.0125], + [-0.0288, -0.0029, 0.0204, ..., 0.0076, 0.0221, -0.0087], + ..., + [ 0.0242, 0.0092, -0.0107, ..., 0.0071, 0.0128, 0.0153], + [ 0.0269, -0.0093, 0.0172, ..., -0.0216, -0.0184, -0.0106], + [ 0.0356, -0.0074, 0.0089, ..., -0.0160, 0.0129, -0.0157]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm.weight': tensor([ 1.2268e-02, 4.5475e-13, -9.5461e-09, -4.1389e-04, 9.3223e-11, + 8.0566e-03, -3.4948e-23, 4.6921e-04, 1.1511e-12, 3.6163e-03, + 1.2293e-07, 2.2583e-02, -3.2131e-08, -2.1118e-02, -1.6764e-08, + 1.5616e-05, 9.6436e-03, -2.1316e-14, 3.3936e-02, 3.7253e-07, + 2.0862e-06, 7.2327e-03, 6.0547e-01, 2.7148e-01, 2.9220e-08, + -3.5048e-05, -2.8871e-08, 1.0610e-05, 3.9648e-01, -4.4250e-03, + 5.0641e-09, 4.5076e-07, 1.3065e-04, 8.6799e-07, -4.6799e-08, + -1.3635e-06, 1.1414e-02, -2.9206e-05, -8.1855e-11, 7.2718e-06, + 1.1362e-07, -8.9264e-04, -3.5858e-04, -9.0599e-06, 7.0190e-03, + -1.1230e-02, -3.6955e-06, 2.2217e-02, -1.7285e-05, -3.5286e-04, + 8.6060e-03, -3.6694e-07, 2.7954e-02, 6.1035e-03, -3.8708e-09, + 1.2817e-02, -1.6093e-06, 1.3590e-05, 2.9541e-02, 4.5776e-03, + 6.4850e-05, -2.0981e-05, -1.7136e-06, -2.7120e-06, 1.4893e-02, + 2.6978e-02, -8.0466e-06, 7.6294e-03, -1.8120e-05, 6.2466e-05, + 2.3804e-02, 2.4319e-05, 1.2008e-12, -1.0347e-04, 1.1222e-07, + -7.9274e-06, 8.6670e-03, 2.6001e-02, 1.1108e-02, 3.0029e-02, + 6.7444e-03, 1.2591e-06, -9.0804e-09, 9.2667e-08, 2.2949e-02, + 1.2970e-03, 3.2959e-02, 1.2070e-06, 1.2329e-02, 2.5868e-05, + -1.8533e-07, 2.0264e-02, 1.6689e-06, -4.3030e-03, -7.7125e-10, + 4.1016e-02, -1.8477e-05, -3.1869e-09, -3.3178e-09, 1.5640e-03, + 2.8849e-05, 2.3438e-02, -2.9907e-02, 9.4771e-06, 8.3008e-03, + -6.9737e-06, 9.7046e-03, -2.8419e-04, 2.0752e-02, -7.5073e-03, + 1.7390e-09, 1.4063e-07, -7.4768e-04, 9.4604e-03, 3.1433e-03, + -1.2219e-05, 2.1606e-02, 2.0752e-02, 2.6733e-02, 3.9935e-06, + 8.1956e-08, 3.5858e-03, -1.4782e-05, 5.3942e-06, 4.3701e-02, + -1.1902e-03, 6.3446e-09, 4.0283e-02], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.v_proj.weight': tensor([[ 0.0166, 0.0034, 0.0182, ..., 0.0041, 0.0141, 0.0089], + [-0.0104, -0.0031, -0.0164, ..., -0.0042, -0.0110, -0.0081], + [-0.0064, -0.0010, -0.0156, ..., -0.0030, -0.0089, -0.0071], + ..., + [ 0.0153, 0.0031, 0.0160, ..., 0.0043, 0.0117, 0.0068], + [ 0.0064, -0.0014, -0.0096, ..., -0.0034, -0.0063, -0.0156], + [ 0.0033, 0.0003, 0.0134, ..., 0.0008, 0.0114, 0.0055]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.output_proj.weight': tensor([[ 0.0142, -0.0118, -0.0109, ..., 0.0007, -0.0051, 0.0023], + [ 0.0047, -0.0027, 0.0011, ..., 0.0074, -0.0015, 0.0056], + [-0.0025, 0.0011, 0.0018, ..., -0.0059, 0.0107, -0.0006], + ..., + [ 0.0019, -0.0024, -0.0018, ..., -0.0025, -0.0046, -0.0016], + [ 0.0008, -0.0038, -0.0059, ..., 0.0011, -0.0030, 0.0095], + [ 0.0073, -0.0010, 0.0008, ..., -0.0151, 0.0167, -0.0133]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img.weight': tensor([[-0.0178, 0.0378, 0.0284, ..., 0.0200, -0.0066, -0.0094], + [ 0.0063, 0.0425, 0.0289, ..., 0.0117, 0.0232, 0.0022], + [-0.0206, 0.0417, -0.0125, ..., -0.0073, -0.0039, 0.0036], + ..., + [-0.0089, -0.0410, 0.0166, ..., -0.0222, 0.0039, 0.0245], + [ 0.0189, 0.0131, -0.0249, ..., 0.0698, -0.0211, -0.0223], + [ 0.0552, -0.0114, 0.0027, ..., -0.0164, -0.0020, 0.0019]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.v_img.weight': tensor([[ 0.0033, 0.0498, 0.0302, ..., 0.0245, -0.0309, -0.0164], + [-0.0143, -0.0074, 0.0315, ..., -0.0432, -0.0027, -0.0151], + [-0.0161, -0.0654, -0.0062, ..., -0.0059, -0.0026, 0.0226], + ..., + [ 0.0056, -0.0231, -0.0255, ..., 0.0038, 0.0295, -0.0057], + [-0.0255, 0.0513, 0.0029, ..., 0.0317, 0.0349, 0.0388], + [ 0.0364, 0.0474, -0.0031, ..., -0.0334, -0.0510, 0.0371]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img.weight': tensor([[ 0.0098, -0.0156, 0.0250, ..., 0.0106, 0.0288, 0.0074], + [-0.0067, -0.0070, -0.0364, ..., -0.0175, 0.0306, -0.0078], + [-0.0051, -0.0195, -0.0012, ..., 0.0410, 0.0732, 0.0040], + ..., + [ 0.0172, 0.0075, 0.0334, ..., -0.0097, 0.0464, 0.0096], + [-0.0242, -0.0233, 0.0297, ..., 0.0072, -0.0193, -0.0140], + [-0.0166, 0.0062, 0.0082, ..., -0.0405, 0.0320, 0.0231]], + dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.blocks.27.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.blocks.27.mlp.layer1.weight': tensor([[-4.6997e-03, 2.7008e-03, 7.6675e-04, ..., -2.5787e-03, + 2.0874e-02, 1.8555e-02], + [ 1.2207e-02, -2.4048e-02, 2.1973e-02, ..., 2.9755e-03, + 8.2493e-05, 4.5166e-03], + [-2.9449e-03, -8.0109e-04, 2.8687e-03, ..., -2.0752e-03, + 2.8198e-02, -2.6093e-03], + ..., + [-2.7161e-03, 3.4714e-04, 1.3184e-02, ..., 7.1716e-03, + -9.1553e-03, -7.7515e-03], + [ 3.9673e-03, -2.7924e-03, -6.2561e-03, ..., 1.2878e-02, + -1.8799e-02, 1.0681e-02], + [-1.3916e-02, 1.3489e-02, 9.1553e-04, ..., -3.2196e-03, + 1.4648e-02, 1.1063e-03]], dtype=torch.bfloat16), 'net.blocks.27.mlp.layer2.weight': tensor([[ 0.0089, 0.0020, 0.0023, ..., -0.0045, -0.0128, -0.0148], + [ 0.0072, 0.0159, -0.0036, ..., -0.0037, 0.0075, -0.0088], + [-0.0021, 0.0211, -0.0021, ..., -0.0014, -0.0059, -0.0302], + ..., + [-0.0070, 0.0112, -0.0095, ..., 0.0097, -0.0016, 0.0120], + [ 0.0014, -0.0096, 0.0023, ..., -0.0005, -0.0077, 0.0070], + [ 0.0005, -0.0070, -0.0140, ..., 0.0026, 0.0006, -0.0070]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.1.weight': tensor([[ 0.0498, 0.0496, 0.0498, ..., 0.0011, 0.0011, 0.0011], + [-0.0044, -0.0052, -0.0062, ..., -0.0010, -0.0012, -0.0011], + [ 0.0457, 0.0457, 0.0459, ..., 0.0011, 0.0011, 0.0012], + ..., + [ 0.0586, 0.0586, 0.0586, ..., 0.0010, 0.0010, 0.0009], + [-0.0640, -0.0625, -0.0615, ..., 0.0016, 0.0014, 0.0015], + [-0.0615, -0.0603, -0.0588, ..., 0.0014, 0.0014, 0.0013]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_self_attn.2.weight': tensor([[-0.0005, 0.0005, -0.0005, ..., -0.0005, -0.0019, -0.0015], + [-0.0007, 0.0005, -0.0007, ..., -0.0006, 0.0027, 0.0021], + [-0.0017, 0.0019, -0.0019, ..., -0.0017, 0.0012, 0.0010], + ..., + [-0.0879, 0.0811, -0.0874, ..., -0.0889, -0.0669, -0.0737], + [-0.0767, 0.0957, -0.0781, ..., -0.0752, -0.1719, -0.1729], + [ 0.0820, -0.0767, 0.0825, ..., 0.0815, 0.0422, 0.0532]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.1.weight': tensor([[-2.6001e-02, -2.5513e-02, -2.4902e-02, ..., 1.3447e-04, + -2.8014e-05, -2.3246e-05], + [-4.9805e-02, -4.8828e-02, -4.7607e-02, ..., 3.2425e-05, + -4.1962e-05, -2.2650e-06], + [-5.4443e-02, -5.4199e-02, -5.2979e-02, ..., -1.0490e-05, + 1.2398e-04, -3.4809e-05], + ..., + [-2.9419e-02, -2.8687e-02, -2.8198e-02, ..., 8.2016e-05, + -3.6955e-05, -3.0398e-05], + [-2.0020e-02, -1.9653e-02, -1.9165e-02, ..., 1.3638e-04, + 5.1260e-05, 2.7061e-05], + [ 6.4697e-03, 6.1951e-03, 5.8899e-03, ..., -5.7697e-05, + -8.9645e-05, 2.3842e-05]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_cross_attn.2.weight': tensor([[-1.5564e-03, -2.4872e-03, -2.0599e-03, ..., -2.3804e-03, + -8.5068e-04, 7.3910e-05], + [ 8.3618e-03, 7.5684e-03, 6.0425e-03, ..., 9.5215e-03, + 7.4463e-03, -6.1035e-03], + [-4.3640e-03, 1.9150e-03, 1.9226e-03, ..., -3.6163e-03, + -4.7607e-03, 4.2419e-03], + ..., + [ 1.1414e-02, -1.3794e-02, -1.9897e-02, ..., 8.1787e-03, + 1.6968e-02, -2.9175e-02], + [ 1.2024e-02, -3.4424e-02, -3.9307e-02, ..., 7.1106e-03, + 1.8921e-02, -3.4180e-02], + [-8.1177e-03, 5.3101e-03, 1.0193e-02, ..., -6.5613e-03, + -1.0742e-02, 1.6968e-02]], dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.1.weight': tensor([[-0.0439, -0.0437, -0.0435, ..., -0.0001, -0.0002, -0.0001], + [ 0.0708, 0.0693, 0.0688, ..., -0.0005, -0.0004, -0.0004], + [-0.0510, -0.0505, -0.0496, ..., 0.0007, 0.0007, 0.0006], + ..., + [-0.0481, -0.0474, -0.0469, ..., 0.0004, 0.0005, 0.0005], + [ 0.0500, 0.0496, 0.0486, ..., -0.0007, -0.0006, -0.0006], + [ 0.0557, 0.0552, 0.0544, ..., -0.0003, -0.0005, -0.0005]], + dtype=torch.bfloat16), 'net.blocks.27.adaln_modulation_mlp.2.weight': tensor([[-8.0585e-05, 9.1934e-04, -9.2697e-04, ..., -4.7874e-04, + 9.1934e-04, 9.2316e-04], + [-4.8828e-04, 9.1934e-04, -8.1253e-04, ..., -7.0190e-04, + 8.4305e-04, 9.0790e-04], + [ 1.5411e-03, -5.6458e-04, 9.4986e-04, ..., 9.6512e-04, + -9.5367e-04, -9.6893e-04], + ..., + [ 3.1006e-02, -4.3701e-02, 4.7607e-02, ..., 4.4434e-02, + -5.1025e-02, -4.1748e-02], + [ 2.6733e-02, -2.9175e-02, 2.8687e-02, ..., 2.7954e-02, + -2.8198e-02, -2.8687e-02], + [-3.1982e-02, 4.5410e-02, -4.5654e-02, ..., -4.3457e-02, + 4.7607e-02, 4.2725e-02]], dtype=torch.bfloat16), 'net.final_layer.linear.weight': tensor([[ 4.8218e-03, -3.4332e-04, -1.7166e-04, ..., -1.2779e-04, + -6.7234e-05, 1.1778e-04], + [ 5.7983e-04, 2.1839e-04, -6.1035e-05, ..., 2.6131e-04, + -4.9973e-04, -2.7180e-05], + [ 5.8746e-04, 1.2207e-03, -1.5259e-04, ..., -1.0490e-04, + 4.3678e-04, -5.0354e-04], + ..., + [-4.4556e-03, 4.0245e-04, 7.3433e-05, ..., -2.0599e-04, + -3.6812e-04, -1.5163e-04], + [-3.9978e-03, 3.0708e-04, 6.1417e-04, ..., -5.5075e-05, + 3.1471e-04, 1.3828e-04], + [ 2.2461e-02, 2.1839e-04, -8.1635e-04, ..., 3.4904e-04, + 8.4877e-05, -3.5286e-04]], dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.1.weight': tensor([[ 0.0300, 0.0299, 0.0299, ..., 0.0033, 0.0031, 0.0030], + [ 0.0091, 0.0094, 0.0095, ..., 0.0023, 0.0023, 0.0024], + [-0.0669, -0.0654, -0.0649, ..., -0.0002, -0.0002, -0.0003], + ..., + [ 0.0986, 0.0962, 0.0933, ..., -0.0028, -0.0027, -0.0028], + [-0.0439, -0.0437, -0.0437, ..., -0.0009, -0.0010, -0.0011], + [ 0.0474, 0.0474, 0.0466, ..., 0.0008, 0.0007, 0.0007]], + dtype=torch.bfloat16), 'net.final_layer.adaln_modulation.2.weight': tensor([[-0.0008, -0.0015, -0.0012, ..., -0.0002, -0.0003, 0.0003], + [-0.0004, -0.0003, 0.0002, ..., 0.0020, 0.0003, -0.0003], + [-0.0008, -0.0020, -0.0019, ..., 0.0149, -0.0006, 0.0005], + ..., + [-0.0067, -0.0040, 0.0266, ..., -0.0791, 0.0101, -0.0111], + [-0.0115, 0.0027, 0.0366, ..., -0.1465, 0.0226, -0.0261], + [-0.0063, -0.0023, 0.0188, ..., -0.0698, 0.0095, -0.0104]], + dtype=torch.bfloat16), 'net.t_embedding_norm.weight': tensor([0.0811, 0.0791, 0.0771, ..., 0.0352, 0.0361, 0.0359], + dtype=torch.bfloat16), 'net.t_embedding_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.img_context_proj.0.weight': tensor([[-0.0133, -0.0056, 0.0145, ..., -0.0072, 0.0211, -0.0275], + [ 0.0078, -0.0057, -0.0234, ..., 0.0084, 0.0071, -0.0166], + [ 0.0225, -0.0251, 0.0052, ..., 0.0078, 0.0236, -0.0046], + ..., + [ 0.0186, 0.0143, -0.0046, ..., 0.0272, -0.0083, -0.0096], + [ 0.0077, 0.0035, 0.0072, ..., -0.0260, 0.0267, 0.0151], + [ 0.0103, -0.0076, 0.0212, ..., 0.0195, -0.0157, 0.0121]], + dtype=torch.bfloat16), 'net.img_context_proj.0.bias': tensor([-0.0087, -0.0270, 0.0251, ..., 0.0096, 0.0049, -0.0209], + dtype=torch.bfloat16), 'net.crossattn_proj.0.weight': tensor([[-0.0067, 0.0204, 0.0132, ..., 0.0005, 0.0051, -0.0074], + [ 0.0013, 0.0009, -0.0017, ..., -0.0007, -0.0008, -0.0018], + [ 0.0095, -0.0031, 0.0057, ..., 0.0049, -0.0006, 0.0079], + ..., + [ 0.0056, 0.0015, 0.0084, ..., -0.0012, 0.0056, 0.0034], + [-0.0093, -0.0184, -0.0043, ..., -0.0065, 0.0041, -0.0125], + [-0.0170, 0.0017, -0.0007, ..., 0.0064, -0.0009, -0.0006]], + dtype=torch.bfloat16), 'net.crossattn_proj.0.bias': tensor([0.8281, 0.7070, 0.8164, ..., 0.6523, 0.8242, 0.8203], + dtype=torch.bfloat16), 'net.control_embedder.3.proj.1.weight': tensor([[-0.0172, 0.0071, 0.0757, ..., -0.0610, 0.0327, -0.0320], + [-0.0278, 0.0029, -0.1128, ..., 0.0272, -0.0361, 0.0583], + [-0.0347, 0.0947, -0.0361, ..., -0.0166, 0.0172, -0.0332], + ..., + [ 0.0125, 0.0007, -0.0100, ..., 0.0219, -0.0376, -0.0036], + [-0.0239, -0.0574, 0.0586, ..., 0.0378, 0.0830, 0.0608], + [ 0.0176, -0.0205, 0.0076, ..., 0.0248, 0.0493, 0.0039]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_proj.weight': tensor([[-4.1199e-03, 5.4932e-03, 6.3782e-03, ..., 2.3346e-03, + -5.4359e-05, 8.1177e-03], + [-9.6893e-04, 8.4305e-04, -8.3923e-04, ..., -8.9111e-03, + -4.5586e-04, 2.2736e-03], + [ 4.7913e-03, -3.5553e-03, 3.2806e-03, ..., -8.0566e-03, + -4.3335e-03, 1.3809e-03], + ..., + [-2.3193e-03, -4.1504e-03, -7.0496e-03, ..., 1.2268e-02, + 4.0588e-03, -8.0566e-03], + [-9.6436e-03, -7.9346e-03, -1.2634e-02, ..., -1.3123e-02, + -5.9204e-03, -6.7139e-04], + [-1.7929e-03, -8.9111e-03, 6.1646e-03, ..., 9.9487e-03, + -4.2915e-04, 1.8005e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_norm.weight': tensor([ 4.7852e-02, -1.6968e-02, 4.7119e-02, 7.8613e-02, -1.3809e-03, + 4.0513e-08, -1.1362e-16, -1.8396e-21, -3.8598e-17, -1.1689e-19, + -5.3204e-21, 3.8257e-24, -6.4851e-22, -1.0727e-24, -2.7504e-23, + 3.6028e-25, -6.9384e-08, -4.4250e-03, -2.9395e-09, -8.8476e-09, + 7.6834e-09, 1.5924e-19, 2.1973e-02, 3.1281e-04, -5.5908e-02, + 4.2578e-01, -1.0889e-01, -3.9062e-03, 1.7822e-02, 8.7891e-02, + 1.0029e-18, 2.1537e-08, 2.6855e-03, -5.1737e-05, -5.5552e-05, + 5.1880e-04, -4.3074e-09, 8.7486e-14, -4.0118e-23, 2.0195e-26, + -7.4216e-27, 8.4703e-22, 3.2558e-21, -1.4343e-03, 1.9775e-02, + -3.8719e-04, -5.4688e-02, -5.0293e-02, 1.7969e-01, 2.9541e-02, + -5.5542e-03, 8.2397e-04, -3.9062e-03, 3.2596e-07, 7.0690e-17, + -7.6762e-22, 5.8981e-17, 1.3897e-22, -2.2334e-22, 9.1382e-27, + 2.7297e-22, -1.3944e-13, 1.9690e-27, 6.0269e-28, 1.2988e-01, + 1.4746e-01, 3.4570e-01, 7.7820e-03, 1.8311e-03, 2.5728e-08, + 7.2858e-17, 3.7481e-20, -1.4162e-18, -9.3512e-19, -7.4115e-20, + -1.3329e-25, 2.2658e-20, -5.5008e-23, 2.5312e-22, 7.6256e-25, + -1.3318e-07, -7.6771e-05, -2.6375e-06, 3.3424e-11, 2.0210e-07, + -1.2559e-15, -2.5635e-03, -1.6022e-03, 3.4570e-01, -1.5625e-02, + 4.1992e-01, 7.9102e-02, -8.3618e-03, 5.3711e-03, -1.2536e-18, + 1.3132e-07, 6.4087e-04, -9.7156e-06, -3.7193e-05, 5.3406e-05, + -7.4971e-08, 1.2546e-14, -2.7303e-25, -2.2877e-24, -4.2662e-27, + -1.5096e-23, 7.9939e-21, 1.2756e-02, 2.8076e-03, -1.2054e-03, + 3.9844e-01, 2.0703e-01, 1.5430e-01, -1.6113e-01, 7.7515e-03, + -4.9805e-02, -3.1128e-03, 1.5087e-07, 2.9707e-17, -5.7439e-21, + 3.4911e-17, 3.6528e-21, 7.3268e-20, 4.3823e-26, 1.4540e-24, + 1.9895e-11, 7.8065e-24, -1.0057e-25], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.self_attn.k_proj.weight': tensor([[-7.0496e-03, 4.8218e-03, -7.4768e-03, ..., -4.2114e-03, + -3.3264e-03, 2.1515e-03], + [-1.9287e-02, -1.4496e-03, 1.9226e-03, ..., 1.4114e-03, + -5.8289e-03, -1.9531e-03], + [-2.1219e-05, -1.2436e-03, 6.7139e-03, ..., 1.1826e-03, + 1.5488e-03, 8.5449e-03], + ..., + [ 2.5330e-03, 2.9602e-03, -2.6550e-03, ..., -5.2261e-04, + -1.1215e-03, 3.8910e-03], + [ 1.1368e-03, 1.0757e-03, -7.6599e-03, ..., -1.0529e-03, + -1.5354e-04, -1.1063e-03], + [ 1.4648e-03, 1.9684e-03, -3.1738e-03, ..., -5.8594e-03, + -2.7466e-04, 1.3962e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.k_norm.weight': tensor([ 9.4727e-02, 5.6641e-02, -5.5664e-02, 7.6660e-02, 1.5335e-03, + -2.1770e-08, -1.1189e-16, 1.7999e-20, -3.8381e-17, -1.6178e-19, + -8.7350e-21, -3.7740e-24, -2.6073e-21, 1.4605e-24, 1.8301e-23, + 6.4300e-25, -6.6590e-08, 4.1504e-03, 1.0477e-09, 8.8476e-09, + 7.5088e-09, -1.1181e-19, 1.6724e-02, 1.3504e-03, -1.5625e-02, + 4.2383e-01, -1.1816e-01, 7.5684e-02, 3.2349e-03, 4.5538e-05, + -1.4704e-18, 6.2864e-08, -2.6093e-03, -5.1498e-05, -5.5313e-05, + -5.4932e-04, -8.7894e-09, -7.1942e-14, 4.0325e-23, 2.8071e-26, + -6.6643e-27, -8.4372e-22, -5.1351e-21, 1.0193e-02, 1.8311e-02, + 2.2125e-04, 8.5449e-02, 1.2598e-01, 2.0801e-01, 1.6211e-01, + 7.1411e-03, 4.9072e-02, -3.7994e-03, -3.8557e-07, -4.6079e-18, + -1.6874e-21, 5.2313e-18, 4.4668e-22, -9.2644e-22, -1.1763e-26, + 2.6635e-22, -2.5757e-13, -2.0397e-26, -9.4663e-28, 8.1177e-03, + 1.4355e-01, 3.4766e-01, -4.5898e-02, 1.6556e-03, 4.7730e-08, + 5.6379e-17, 3.5999e-20, 1.6263e-18, 9.1480e-19, -7.3692e-20, + -6.8259e-26, 2.0964e-20, 5.5421e-23, -2.6304e-22, 7.8194e-25, + -1.6857e-07, -8.5354e-05, -2.6226e-06, 7.9012e-12, 1.9558e-07, + 1.2976e-15, -1.5564e-03, -3.9291e-04, 3.4570e-01, -1.6479e-02, + 4.1992e-01, 5.7983e-03, 1.7578e-02, 8.7402e-02, -1.7872e-19, + 6.9849e-08, 4.4632e-04, 1.0669e-05, -3.7432e-05, 4.8637e-05, + -7.9628e-08, -1.2379e-14, -1.7933e-25, -2.0550e-24, 4.2914e-27, + 1.5303e-23, -7.9409e-21, 8.2397e-03, 1.1292e-02, 7.4387e-04, + 4.0039e-01, 1.8359e-01, 1.4355e-01, 1.5625e-02, 7.4768e-03, + -1.1597e-03, -3.6621e-03, -8.1491e-08, -3.8164e-17, 5.7175e-21, + -3.7513e-17, -3.4940e-21, -7.1998e-20, 2.3426e-26, -2.2101e-24, + 1.9895e-11, 7.8065e-24, -1.0138e-25], dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.self_attn.v_proj.weight': tensor([[-0.0040, 0.0033, -0.0051, ..., -0.0042, 0.0036, -0.0044], + [ 0.0073, -0.0018, 0.0019, ..., -0.0043, -0.0013, -0.0023], + [ 0.0152, 0.0034, -0.0048, ..., 0.0084, -0.0022, -0.0011], + ..., + [ 0.0050, 0.0029, -0.0007, ..., -0.0044, 0.0009, 0.0006], + [ 0.0009, 0.0044, -0.0021, ..., -0.0018, 0.0045, -0.0052], + [ 0.0008, -0.0037, 0.0006, ..., -0.0032, -0.0033, -0.0094]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.self_attn.output_proj.weight': tensor([[ 0.0129, -0.0291, -0.0254, ..., 0.0038, 0.0062, 0.0021], + [ 0.0141, -0.0053, 0.0117, ..., 0.0043, 0.0077, 0.0028], + [-0.0046, 0.0078, 0.0023, ..., 0.0040, 0.0029, -0.0049], + ..., + [ 0.0094, -0.0002, -0.0016, ..., -0.0038, 0.0046, -0.0112], + [-0.0047, 0.0088, 0.0038, ..., -0.0010, 0.0015, 0.0010], + [-0.0056, 0.0040, -0.0090, ..., 0.0036, 0.0044, -0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_proj.weight': tensor([[-0.0025, -0.0159, 0.0013, ..., 0.0058, -0.0039, 0.0007], + [ 0.0072, 0.0210, 0.0029, ..., -0.0058, 0.0050, -0.0034], + [ 0.0085, 0.0194, 0.0049, ..., -0.0053, 0.0130, -0.0042], + ..., + [ 0.0072, 0.0052, 0.0026, ..., 0.0013, -0.0083, 0.0012], + [-0.0029, -0.0007, -0.0029, ..., -0.0027, 0.0109, -0.0044], + [ 0.0002, -0.0038, 0.0008, ..., -0.0022, 0.0015, 0.0057]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_norm.weight': tensor([ 2.0695e-04, 1.8677e-02, 2.3192e-11, 5.6458e-04, 3.6049e-04, + -9.7156e-06, -6.0499e-06, 2.2125e-03, 3.6377e-02, 2.6611e-02, + 1.4305e-04, 3.8330e-02, 5.5859e-01, 3.2196e-03, 8.4839e-03, + 1.1981e-05, 4.2969e-01, 6.8188e-05, 1.9360e-04, 1.1292e-02, + 3.8528e-04, 1.5640e-03, 9.1309e-02, 1.1902e-02, 4.0283e-03, + 6.3171e-03, 2.3682e-02, 3.0696e-06, 9.1553e-03, 2.3651e-04, + 3.6133e-01, 1.9165e-02, 6.0120e-03, 3.8281e-01, 5.3125e-01, + 9.4986e-04, 8.8882e-04, 2.5879e-02, 1.4258e-01, 8.6060e-03, + 1.6113e-02, 2.0142e-03, -4.4890e-07, 3.3875e-03, 1.8921e-03, + 4.7119e-02, 9.3994e-03, 4.0770e-05, 2.3041e-03, 2.0218e-04, + 1.9932e-04, 1.8552e-06, 1.4156e-06, -1.7762e-05, -3.5400e-03, + 1.3672e-02, 2.6367e-02, 1.3275e-03, 1.4954e-02, 1.4526e-02, + -7.0572e-05, -2.9057e-07, 2.9945e-04, 8.1253e-04, 1.1658e-02, + 4.3640e-03, 1.5640e-03, 2.2070e-01, 1.0803e-02, 2.4780e-02, + 3.9258e-01, 1.5259e-05, 4.6143e-02, 2.0020e-02, 3.7842e-03, + -1.9550e-04, 4.4556e-03, 6.7139e-03, 6.1035e-03, 9.2697e-04, + 9.8267e-03, 1.0071e-02, 1.6016e-01, 1.7929e-03, 2.3127e-05, + 1.1719e-02, -1.5945e-03, 1.1265e-05, 8.9111e-03, 4.7852e-01, + 5.0049e-03, 6.5231e-04, 1.4282e-02, 6.4697e-03, 2.5630e-06, + 4.1406e-01, 1.6928e-05, 1.7929e-04, -1.6964e-13, 6.5613e-03, + 3.2617e-01, -2.6524e-06, -4.1199e-03, 4.0283e-03, 3.4912e-02, + 6.1512e-05, -2.4199e-05, 1.8477e-05, 1.9775e-02, 1.0132e-02, + 1.3504e-03, -3.4332e-05, 9.3994e-03, 4.8218e-03, 2.2430e-03, + 7.1335e-04, 2.0874e-02, 1.7471e-03, -9.5367e-04, 5.3101e-03, + 4.9353e-05, 4.5471e-03, 9.1934e-04, 1.2054e-03, 1.5503e-02, + 1.4210e-04, 2.7161e-03, 1.8501e-04], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_proj.weight': tensor([[-0.0022, 0.0007, -0.0010, ..., -0.0041, 0.0014, -0.0015], + [-0.0013, -0.0021, -0.0042, ..., -0.0038, -0.0054, -0.0001], + [ 0.0043, -0.0015, -0.0052, ..., -0.0038, -0.0040, -0.0036], + ..., + [ 0.0038, -0.0025, 0.0025, ..., 0.0010, 0.0006, -0.0012], + [ 0.0008, 0.0005, -0.0023, ..., -0.0032, 0.0006, 0.0013], + [ 0.0017, 0.0019, -0.0031, ..., -0.0045, -0.0006, 0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_norm.weight': tensor([ 1.6499e-04, 1.8677e-02, -2.0009e-11, 2.2125e-04, 3.4523e-04, + 1.2815e-05, 6.0499e-06, 2.1667e-03, 3.5645e-02, 2.6611e-02, + 1.5354e-04, 3.8574e-02, 5.6250e-01, 3.1738e-03, 8.0566e-03, + 9.1195e-06, 4.2969e-01, 6.8188e-05, -1.1015e-04, 1.1230e-02, + 6.5994e-04, 1.5945e-03, 9.1309e-02, 1.2024e-02, 3.8757e-03, + 6.3171e-03, 2.3438e-02, 2.9057e-06, 8.5449e-03, 2.3651e-04, + 3.6133e-01, 1.8921e-02, 5.9509e-03, 3.8281e-01, 5.3125e-01, + 8.7738e-04, 1.0223e-03, 2.6123e-02, 1.4355e-01, 8.7280e-03, + 1.6113e-02, 2.0142e-03, -4.8429e-07, 3.1891e-03, 1.3733e-03, + 4.6875e-02, 9.3384e-03, -1.4901e-05, 2.3193e-03, -7.9155e-05, + 2.0218e-04, 1.2815e-06, 1.5944e-06, 2.4915e-05, 3.9673e-03, + 1.3550e-02, 2.6245e-02, 1.3504e-03, 1.5259e-02, 1.4587e-02, + -7.7248e-05, -5.5507e-07, -3.1471e-04, 7.4005e-04, 1.1719e-02, + 4.3335e-03, 1.5640e-03, 2.2070e-01, 1.0742e-02, 2.4902e-02, + 3.9258e-01, 1.5259e-05, 4.4434e-02, 2.0020e-02, 3.7994e-03, + 1.9455e-04, 3.6926e-03, 6.6833e-03, 6.1646e-03, 9.1171e-04, + 9.7656e-03, 1.0071e-02, 1.6016e-01, 1.5564e-03, -1.7524e-05, + 1.1719e-02, 1.9379e-03, 1.1861e-05, 8.9722e-03, 4.7852e-01, + 4.9744e-03, 6.1798e-04, 1.4099e-02, 6.4697e-03, 2.2501e-06, + 4.1406e-01, 4.2021e-06, 7.7724e-05, 2.2116e-13, 6.5613e-03, + 3.2812e-01, 2.6673e-06, 4.2114e-03, 4.0283e-03, 3.4912e-02, + 6.2943e-05, 3.9101e-05, 1.8835e-05, 1.9165e-02, 1.0193e-02, + 1.5945e-03, 3.5048e-05, 9.3384e-03, 4.7607e-03, 2.6093e-03, + -6.7139e-04, 2.0874e-02, 1.8539e-03, 1.2131e-03, 5.3101e-03, + 3.6955e-05, 4.6997e-03, 9.1171e-04, 1.1749e-03, 1.5076e-02, + -1.4114e-04, 2.8076e-03, -5.2452e-05], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.v_proj.weight': tensor([[-7.5989e-03, -1.1215e-03, 1.2329e-02, ..., -7.8964e-04, + 5.2490e-03, -1.0742e-02], + [-2.1729e-02, -2.2278e-03, 5.7068e-03, ..., -3.0060e-03, + -8.9111e-03, -2.3956e-03], + [-9.3994e-03, -1.8768e-03, 4.5776e-03, ..., -4.2114e-03, + 1.4771e-02, -1.2085e-02], + ..., + [ 1.0010e-02, -5.7220e-04, 4.4346e-05, ..., -8.2779e-04, + 3.7384e-03, 1.9897e-02], + [-7.6294e-03, 1.6708e-03, 1.0803e-02, ..., 1.3046e-03, + -1.7334e-02, -1.2360e-03], + [-1.6846e-02, -1.8082e-03, 8.1177e-03, ..., -1.2665e-03, + -7.4768e-03, -2.2583e-02]], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.output_proj.weight': tensor([[-0.0013, -0.0039, 0.0039, ..., 0.0077, 0.0042, 0.0098], + [ 0.0042, 0.0090, -0.0073, ..., 0.0137, 0.0020, -0.0074], + [ 0.0077, -0.0078, -0.0010, ..., 0.0038, -0.0059, -0.0006], + ..., + [-0.0054, -0.0056, -0.0043, ..., -0.0075, 0.0052, -0.0055], + [ 0.0081, 0.0005, 0.0091, ..., -0.0078, -0.0079, -0.0040], + [ 0.0175, -0.0094, -0.0039, ..., -0.0023, -0.0044, 0.0064]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_img.weight': tensor([[ 0.0078, 0.0121, 0.0018, ..., -0.0161, 0.0175, -0.0481], + [ 0.0165, -0.0464, 0.0211, ..., 0.0074, 0.0249, 0.0291], + [-0.0095, 0.0080, -0.0121, ..., -0.0342, -0.0327, 0.0065], + ..., + [ 0.0051, -0.0067, 0.0038, ..., 0.0084, 0.0286, -0.0165], + [ 0.0078, -0.0067, -0.0101, ..., -0.0095, 0.0140, -0.0309], + [ 0.0242, -0.0101, 0.0255, ..., -0.0227, 0.0075, -0.0148]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.v_img.weight': tensor([[-0.0649, 0.0383, 0.0305, ..., -0.0220, 0.0247, 0.0018], + [ 0.0041, 0.0120, 0.0405, ..., 0.0126, -0.0410, 0.0361], + [-0.0089, 0.0115, -0.0080, ..., -0.0056, -0.0057, 0.0291], + ..., + [-0.0256, -0.0281, 0.0260, ..., -0.0342, 0.0045, 0.0052], + [-0.0160, 0.0042, 0.0036, ..., 0.0133, -0.0173, 0.0024], + [ 0.0038, -0.0209, 0.0442, ..., -0.0325, 0.0231, -0.0099]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img.weight': tensor([[-0.0510, 0.0054, -0.0364, ..., -0.0225, -0.0145, -0.0052], + [-0.0157, -0.0052, -0.0364, ..., 0.0095, 0.0111, -0.0023], + [ 0.0037, 0.0615, 0.0018, ..., -0.0205, -0.0306, -0.0115], + ..., + [-0.0043, 0.0042, -0.0099, ..., -0.0211, -0.0044, 0.0288], + [ 0.0266, -0.0117, 0.0047, ..., -0.0420, -0.0282, -0.0014], + [ 0.0286, 0.0325, 0.0094, ..., -0.0309, -0.0168, 0.0359]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.0.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.0.mlp.layer1.weight': tensor([[ 0.0036, -0.0023, -0.0013, ..., -0.0089, 0.0038, 0.0076], + [-0.0030, 0.0058, -0.0032, ..., -0.0057, 0.0044, 0.0010], + [-0.0087, 0.0012, -0.0061, ..., -0.0046, 0.0040, 0.0060], + ..., + [-0.0016, -0.0054, 0.0029, ..., -0.0075, -0.0067, 0.0027], + [-0.0028, 0.0018, -0.0085, ..., 0.0128, 0.0232, 0.0038], + [ 0.0034, 0.0015, -0.0031, ..., 0.0009, -0.0068, 0.0004]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.mlp.layer2.weight': tensor([[-2.4109e-03, 7.9346e-03, -9.7656e-03, ..., -1.1215e-03, + 2.0599e-03, 1.2268e-02], + [-1.8005e-03, 6.0425e-03, 3.6011e-03, ..., -1.6846e-02, + -2.3071e-02, 4.7607e-03], + [-7.0035e-06, -8.9722e-03, -4.5166e-03, ..., -5.1575e-03, + -7.6294e-03, -1.2573e-02], + ..., + [ 2.1515e-03, 1.1902e-03, -9.5215e-03, ..., 2.8381e-03, + 8.3008e-03, -7.6904e-03], + [-1.0437e-02, 5.0354e-03, 6.3324e-04, ..., 1.0742e-02, + 3.3447e-02, 5.6152e-03], + [-3.8300e-03, 1.1230e-02, 5.2490e-03, ..., 7.4768e-03, + 5.3406e-03, 2.8610e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_self_attn.1.weight': tensor([[-0.0476, -0.0464, -0.0449, ..., 0.0027, 0.0028, 0.0029], + [ 0.0457, 0.0447, 0.0435, ..., -0.0021, -0.0022, -0.0023], + [ 0.0454, 0.0449, 0.0439, ..., -0.0023, -0.0024, -0.0024], + ..., + [ 0.0479, 0.0471, 0.0454, ..., -0.0021, -0.0022, -0.0022], + [ 0.0471, 0.0459, 0.0444, ..., -0.0020, -0.0020, -0.0019], + [ 0.0471, 0.0461, 0.0449, ..., -0.0027, -0.0029, -0.0028]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_self_attn.2.weight': tensor([[-0.0021, 0.0021, 0.0022, ..., 0.0049, 0.0047, 0.0020], + [-0.0017, 0.0016, 0.0016, ..., 0.0049, 0.0052, 0.0017], + [-0.0059, 0.0059, 0.0054, ..., 0.0074, 0.0082, 0.0057], + ..., + [-0.0452, 0.0525, 0.0515, ..., 0.0515, 0.0488, 0.0452], + [ 0.0047, 0.0081, 0.0080, ..., -0.0068, -0.0065, -0.0046], + [ 0.0081, -0.0093, -0.0092, ..., -0.0067, -0.0059, -0.0082]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_cross_attn.1.weight': tensor([[ 5.5847e-03, 5.9509e-03, 7.9346e-03, ..., -6.8665e-04, + -6.1417e-04, -7.8201e-04], + [-3.8086e-02, -3.6621e-02, -3.5889e-02, ..., 3.4332e-04, + 3.5286e-04, 3.6049e-04], + [ 4.2725e-02, 4.1016e-02, 4.0527e-02, ..., -1.5259e-05, + -5.3346e-06, -2.0266e-05], + ..., + [ 3.7842e-02, 3.7109e-02, 3.6377e-02, ..., -4.2152e-04, + -3.5667e-04, -3.9291e-04], + [ 4.9561e-02, 4.8096e-02, 4.7363e-02, ..., -7.9632e-05, + -4.7386e-06, -6.6757e-05], + [ 3.8574e-02, 3.7354e-02, 3.6865e-02, ..., 6.9141e-05, + 1.4782e-04, 9.6321e-05]], dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_cross_attn.2.weight': tensor([[-0.0070, -0.0059, 0.0048, ..., 0.0062, 0.0064, 0.0053], + [-0.0058, -0.0038, -0.0053, ..., 0.0038, 0.0035, 0.0025], + [-0.0043, -0.0044, -0.0085, ..., 0.0043, -0.0074, -0.0071], + ..., + [-0.0078, -0.0134, 0.0259, ..., 0.0132, 0.0273, 0.0216], + [ 0.0114, 0.0058, 0.0103, ..., -0.0056, 0.0052, 0.0107], + [-0.0127, 0.0037, -0.0052, ..., -0.0037, -0.0037, -0.0052]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_mlp.1.weight': tensor([[-0.0050, -0.0051, -0.0050, ..., 0.0005, 0.0007, 0.0007], + [ 0.0530, 0.0510, 0.0498, ..., -0.0020, -0.0024, -0.0024], + [ 0.0444, 0.0432, 0.0422, ..., -0.0033, -0.0036, -0.0036], + ..., + [ 0.0894, 0.0869, 0.0859, ..., -0.0005, -0.0007, -0.0007], + [ 0.0050, 0.0049, 0.0050, ..., -0.0005, -0.0006, -0.0005], + [ 0.0640, 0.0625, 0.0605, ..., -0.0010, -0.0010, -0.0010]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.adaln_modulation_mlp.2.weight': tensor([[-0.0010, 0.0074, 0.0080, ..., 0.0094, 0.0010, 0.0070], + [ 0.0007, 0.0026, 0.0031, ..., 0.0031, -0.0007, 0.0008], + [ 0.0004, 0.0029, 0.0034, ..., 0.0034, -0.0004, 0.0034], + ..., + [ 0.0072, 0.0603, 0.0820, ..., 0.0562, -0.0073, 0.0408], + [ 0.0059, 0.0070, 0.0063, ..., 0.0101, -0.0056, 0.0061], + [-0.0122, -0.0049, -0.0046, ..., -0.0051, 0.0120, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.before_proj.weight': tensor([[ 1.0431e-05, 1.1215e-03, 9.0027e-04, ..., -2.1820e-03, + 2.7657e-04, 3.4027e-03], + [ 3.8147e-03, 5.9509e-04, 1.0681e-02, ..., 1.2207e-02, + -7.5073e-03, -2.5635e-03], + [-1.1215e-03, -3.4637e-03, 2.3346e-03, ..., 1.5030e-03, + -9.9945e-04, -7.0190e-04], + ..., + [ 5.7373e-03, 2.3193e-03, -4.4060e-04, ..., -1.9684e-03, + -1.2207e-04, -5.2490e-03], + [ 2.5635e-03, -4.5166e-03, -4.9133e-03, ..., -1.8692e-03, + 4.3030e-03, -1.2512e-03], + [ 4.6387e-03, 3.4142e-04, -9.4604e-04, ..., 2.0905e-03, + -4.0588e-03, -5.6839e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.0.before_proj.bias': tensor([-0.0016, 0.0080, -0.0006, ..., -0.0053, 0.0027, 0.0032], + dtype=torch.bfloat16), 'net.control_blocks_3.0.after_proj.weight': tensor([[-0.0035, 0.0068, 0.0042, ..., -0.0038, 0.0090, 0.0054], + [-0.0079, -0.0034, 0.0029, ..., -0.0081, 0.0088, -0.0048], + [ 0.0004, -0.0017, -0.0006, ..., -0.0027, 0.0028, -0.0023], + ..., + [ 0.0025, -0.0022, 0.0061, ..., 0.0028, 0.0021, -0.0047], + [ 0.0118, -0.0138, -0.0078, ..., -0.0090, -0.0025, 0.0232], + [ 0.0028, 0.0028, 0.0025, ..., 0.0020, -0.0011, 0.0047]], + dtype=torch.bfloat16), 'net.control_blocks_3.0.after_proj.bias': tensor([ 0.0344, -0.0278, 0.0008, ..., -0.0009, -0.0157, 0.0364], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_proj.weight': tensor([[-0.0245, -0.0016, -0.0092, ..., -0.0034, 0.0038, -0.0042], + [-0.0108, -0.0238, -0.0061, ..., -0.0143, -0.0064, -0.0309], + [ 0.0024, 0.0037, -0.0109, ..., 0.0135, 0.0092, -0.0177], + ..., + [ 0.0058, 0.0115, 0.0014, ..., 0.0066, -0.0047, -0.0063], + [ 0.0062, 0.0038, 0.0052, ..., 0.0110, 0.0022, 0.0024], + [-0.0030, -0.0042, 0.0009, ..., -0.0003, -0.0014, 0.0046]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_norm.weight': tensor([ 1.6211e-01, 5.1953e-01, 9.4727e-02, 6.3672e-01, 5.7031e-01, + 4.7265e-08, -2.4872e-03, -2.4011e-10, 2.0339e-13, -1.2306e-28, + 1.0193e-02, 3.3881e-19, -1.1181e-19, -2.1546e-33, 3.7748e-15, + -2.8407e-33, 1.8066e-01, -9.9182e-04, 4.8065e-04, 7.3547e-03, + 1.1921e-05, 5.1117e-04, 1.9727e-01, 7.4219e-01, 7.6172e-01, + 3.8867e-01, -5.8350e-02, 4.3359e-01, -1.3428e-02, 1.6602e-02, + -2.4128e-04, 1.0991e-14, -1.3123e-02, 6.1989e-06, 5.6458e-04, + -3.1433e-03, 3.5390e-08, 2.6090e-15, 4.0531e-05, 4.1567e-13, + 8.4128e-12, -3.3329e-29, 1.7975e-07, 8.5156e-01, 2.4707e-01, + 2.4121e-01, 2.6562e-01, 7.4609e-01, 1.8555e-01, -4.4189e-02, + 1.1902e-02, 1.8359e-01, 1.4305e-04, -1.0315e-02, -1.3924e-04, + 8.0945e-11, 1.6846e-02, -6.3109e-29, 2.7008e-03, -9.2699e-18, + 9.9487e-03, 1.9560e-34, -1.4572e-15, -1.5247e-26, 3.1445e-01, + 1.9922e-01, 4.9414e-01, 1.5527e-01, 9.0332e-02, 2.5495e-08, + 1.1719e-02, 8.4401e-10, 3.4106e-13, -9.5526e-31, -6.5308e-03, + 1.3485e-18, 1.7047e-20, -3.7074e-33, -2.8644e-14, -3.8519e-34, + -5.4932e-03, -2.9206e-05, -1.2131e-03, 2.9755e-03, 6.2943e-05, + 4.2725e-02, 8.1641e-01, 2.4316e-01, 2.2852e-01, 7.5000e-01, + 2.5586e-01, 2.5391e-01, -1.4893e-02, 4.1809e-03, -1.3306e-02, + 2.4203e-14, 1.0193e-02, 2.1338e-05, 3.3140e-05, 3.3417e-03, + 1.4043e-09, -5.7554e-13, 2.3651e-03, -8.5348e-16, 2.1316e-14, + 2.6770e-32, -1.9222e-06, 2.1582e-01, 7.8125e-01, 7.2656e-01, + 7.3438e-01, 2.1973e-01, 3.7305e-01, -5.6076e-04, 2.4414e-01, + 1.1353e-02, 2.4986e-04, -8.4839e-03, -1.3428e-03, 7.3123e-10, + 2.1606e-02, -5.9165e-29, 3.3875e-03, -1.1350e-19, -6.1035e-04, + 1.2488e-34, -9.4369e-15, -1.3545e-23], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.self_attn.k_proj.weight': tensor([[ 0.0018, -0.0083, 0.0069, ..., -0.0095, -0.0078, -0.0113], + [ 0.0317, -0.0110, 0.0075, ..., -0.0096, -0.0066, -0.0208], + [ 0.0042, 0.0084, 0.0037, ..., 0.0049, 0.0028, 0.0038], + ..., + [ 0.0018, -0.0026, -0.0050, ..., -0.0085, -0.0002, 0.0064], + [-0.0094, 0.0029, 0.0021, ..., 0.0157, 0.0093, -0.0103], + [ 0.0045, -0.0045, -0.0019, ..., -0.0126, -0.0001, 0.0065]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.k_norm.weight': tensor([ 2.1973e-01, 5.3906e-01, -8.6914e-02, 6.4844e-01, 5.7422e-01, + -3.9581e-08, -8.8501e-03, 4.4929e-10, -1.4477e-13, -5.2854e-29, + -1.1414e-02, -5.6243e-19, -1.0927e-19, -5.6273e-34, 3.0809e-15, + -9.8704e-34, 1.7969e-01, -9.8419e-04, -4.8065e-04, 4.8584e-02, + 1.1861e-05, 7.0095e-05, 2.0508e-01, 7.5000e-01, 7.6562e-01, + 1.4746e-01, 1.9629e-01, 4.9414e-01, -1.1292e-02, 1.6479e-02, + 8.1787e-03, 1.7097e-14, 1.5259e-02, 3.3230e-06, 5.8365e-04, + 3.7842e-02, 3.3993e-08, 5.4401e-15, 4.0293e-05, 4.2100e-13, + -8.4697e-12, -3.3921e-29, -6.0350e-07, 8.4766e-01, 2.3438e-01, + 2.1210e-03, 6.4941e-02, 7.4609e-01, 9.1797e-02, 3.6774e-03, + -2.4609e-01, 7.7637e-02, 1.4019e-04, -9.6436e-03, -1.0300e-04, + 1.2915e-10, 1.7578e-02, -6.1531e-29, -2.4567e-03, -7.9689e-18, + 9.7656e-03, 7.8993e-35, -4.8017e-15, 5.9373e-26, 3.0469e-01, + 2.3535e-01, 4.9805e-01, 1.2988e-01, -9.4238e-02, -3.0734e-08, + -7.0496e-03, 3.2742e-10, 2.4691e-13, 1.4298e-29, 1.6403e-03, + 1.3891e-18, 1.8423e-20, -6.9213e-34, -2.8200e-14, -8.5764e-35, + 4.5776e-03, -4.0293e-05, -1.2054e-03, 2.8076e-03, 6.2466e-05, + 4.2969e-02, 8.1250e-01, 1.2988e-01, 1.7969e-01, 7.6953e-01, + 2.2656e-01, 2.3047e-01, 1.6235e-02, -3.2501e-03, 9.4604e-03, + -2.2760e-14, 5.4626e-03, -2.2054e-05, -4.7386e-06, 3.1494e-02, + 1.3461e-09, -5.2935e-13, -2.4109e-03, -2.1927e-15, 9.3814e-15, + 5.4696e-32, -1.9222e-06, 2.5781e-01, 7.8516e-01, 7.3047e-01, + 7.4219e-01, -5.1270e-02, 4.0625e-01, -4.3701e-02, 5.5542e-03, + -1.5527e-01, 2.5558e-04, 9.0332e-03, -1.3428e-03, -7.2396e-10, + -1.6846e-02, -6.0348e-29, -3.0823e-03, -1.8465e-19, -5.7220e-04, + 9.7049e-35, -8.5487e-15, -1.1529e-23], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.self_attn.v_proj.weight': tensor([[ 5.7068e-03, 4.6387e-03, 6.9275e-03, ..., -1.5198e-02, + 3.6621e-03, 9.0599e-05], + [ 6.2561e-03, -3.5553e-03, 1.4343e-02, ..., -2.3193e-03, + -2.3956e-03, 3.4523e-04], + [ 4.8828e-03, 4.3030e-03, 3.4790e-03, ..., -6.3782e-03, + 5.6763e-03, -7.4463e-03], + ..., + [ 8.7891e-03, -1.0834e-03, 1.0437e-02, ..., 8.3618e-03, + 3.5248e-03, -1.6724e-02], + [-6.5308e-03, 4.5166e-03, 6.0425e-03, ..., 3.4180e-02, + -2.7924e-03, 5.6458e-03], + [-3.3875e-03, 1.0757e-03, -2.8076e-03, ..., 1.3245e-02, + -4.1199e-03, 4.6692e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.self_attn.output_proj.weight': tensor([[-5.7373e-03, 7.4768e-03, 7.9155e-05, ..., 1.1292e-02, + -1.1719e-02, 7.5073e-03], + [ 1.1047e-02, -3.5553e-03, 1.7578e-02, ..., 4.7913e-03, + -1.2207e-02, 9.5215e-03], + [ 1.9531e-03, 2.9564e-04, 2.3499e-03, ..., 7.7820e-04, + 1.1841e-02, 4.6692e-03], + ..., + [ 1.1047e-02, 4.1504e-03, 1.1749e-03, ..., -4.1809e-03, + -7.9956e-03, -7.2937e-03], + [-7.1106e-03, 8.9722e-03, 4.9438e-03, ..., 4.0588e-03, + -3.7079e-03, 3.6926e-03], + [-1.1475e-02, 9.1553e-03, 1.1597e-02, ..., -2.4719e-03, + 8.1177e-03, -6.1340e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_proj.weight': tensor([[-0.0058, 0.0116, 0.0141, ..., -0.0040, 0.0054, 0.0082], + [ 0.0051, -0.0135, -0.0170, ..., 0.0056, -0.0195, -0.0311], + [ 0.0070, 0.0134, -0.0043, ..., -0.0032, 0.0034, -0.0074], + ..., + [ 0.0020, 0.0212, -0.0030, ..., 0.0063, 0.0024, -0.0049], + [-0.0104, 0.0107, 0.0090, ..., -0.0048, 0.0112, 0.0060], + [ 0.0151, -0.0081, -0.0079, ..., -0.0070, -0.0042, -0.0078]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_norm.weight': tensor([ 1.9775e-02, 4.4727e-01, 2.8320e-02, 4.8578e-06, 3.5048e-05, + 3.2507e-13, 2.3651e-03, 5.1172e-01, 2.6855e-03, 3.7695e-01, + 3.6133e-02, 1.2512e-02, 6.3171e-03, 1.2695e-02, 2.7466e-02, + 1.5381e-02, 1.2207e-02, 7.8201e-05, 2.8163e-06, 3.7109e-02, + 2.5391e-01, 2.5879e-02, 2.4915e-05, 2.4414e-02, 9.4604e-03, + 5.1737e-05, 4.1962e-05, 4.5166e-02, 1.3542e-04, 3.8281e-01, + -4.8399e-05, -1.2684e-04, 5.1880e-03, 8.8215e-05, 2.7008e-03, + 3.2806e-03, 1.9651e-07, 4.1211e-01, 1.0315e-02, 5.4932e-03, + 6.9336e-02, 1.2268e-02, 1.6846e-02, 9.5844e-05, 1.1826e-03, + 1.4526e-02, 3.8910e-03, 2.2217e-02, 2.2461e-02, 1.9775e-02, + 1.1414e-02, 3.8281e-01, 1.2360e-03, 1.9653e-02, 2.6489e-02, + 3.3875e-03, 1.0205e-01, 1.0132e-02, 2.3926e-02, 1.2573e-02, + 1.6975e-04, 5.3101e-03, 1.2878e-02, -1.7738e-04, 5.6152e-03, + 5.5542e-03, -1.0133e-05, 3.3188e-04, 1.6235e-02, 7.9590e-02, + 6.0425e-03, 8.1177e-03, 6.0547e-02, 4.1260e-02, 6.9275e-03, + 5.2185e-03, 7.6771e-05, 2.5879e-02, 3.2227e-02, 2.6489e-02, + 3.9673e-03, 2.0508e-02, 3.1738e-02, -4.5868e-08, 1.9653e-02, + 1.3733e-02, 2.7895e-05, 4.5410e-02, 1.1475e-02, 3.4943e-03, + 1.3916e-02, 7.1335e-04, 5.0049e-03, 2.3193e-02, 1.2878e-02, + -1.3256e-04, 1.3245e-02, 4.0283e-02, 2.0264e-02, 4.8096e-02, + 3.9551e-02, 3.1128e-02, 1.7944e-02, 3.3140e-05, -4.7493e-04, + 3.2616e-04, 1.0864e-02, 2.0264e-02, 1.8799e-02, 1.9287e-02, + 3.7384e-03, 1.3977e-02, 3.6865e-02, 4.9414e-01, -1.6391e-06, + 2.5269e-02, 6.9580e-03, 1.5991e-02, 1.4572e-03, 1.1182e-01, + 5.0068e-06, 2.2411e-05, 3.4571e-05, 1.1475e-02, 1.5015e-02, + 1.5625e-02, 2.3804e-02, 3.1494e-02], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_proj.weight': tensor([[-0.0017, -0.0033, 0.0020, ..., -0.0039, 0.0030, 0.0041], + [ 0.0123, -0.0035, 0.0190, ..., -0.0018, 0.0317, 0.0050], + [ 0.0035, 0.0036, 0.0027, ..., 0.0023, 0.0099, -0.0018], + ..., + [-0.0064, -0.0049, -0.0075, ..., -0.0028, 0.0013, 0.0076], + [-0.0078, -0.0053, -0.0181, ..., -0.0019, 0.0045, 0.0051], + [-0.0024, -0.0004, -0.0070, ..., 0.0022, 0.0089, -0.0009]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_norm.weight': tensor([ 2.0142e-02, 4.4727e-01, 2.8198e-02, 4.8876e-06, 3.5048e-05, + -3.1086e-13, 2.3651e-03, 5.1172e-01, 2.7618e-03, 3.7695e-01, + 3.7109e-02, 1.2512e-02, 6.2866e-03, 1.2817e-02, 2.7588e-02, + 1.5259e-02, 1.2451e-02, 7.7248e-05, 3.7402e-06, 3.7354e-02, + 2.5391e-01, 2.5146e-02, 2.4319e-05, 2.4658e-02, 9.5215e-03, + 5.0783e-05, 4.4346e-05, 4.4678e-02, 1.3733e-04, 3.8477e-01, + -4.8637e-05, -1.2684e-04, 5.0659e-03, 8.7738e-05, 2.7313e-03, + 3.3112e-03, 2.0582e-07, 4.1211e-01, 1.0315e-02, 5.4626e-03, + 6.8359e-02, 1.2207e-02, 1.6724e-02, 9.5367e-05, 1.1063e-03, + 1.4526e-02, 3.9368e-03, 2.1851e-02, 2.2461e-02, 2.0020e-02, + 1.1475e-02, 3.8281e-01, 1.1444e-03, 1.9287e-02, 2.6123e-02, + 3.3112e-03, 1.0205e-01, 1.0132e-02, 2.3438e-02, 1.2512e-02, + 1.6785e-04, 5.3406e-03, 1.3000e-02, -1.7738e-04, 5.6152e-03, + 5.4626e-03, -8.5235e-06, 3.1662e-04, 1.6357e-02, 7.9590e-02, + 6.0120e-03, 8.1177e-03, 6.0059e-02, 4.1260e-02, 6.9885e-03, + 5.3406e-03, 7.7248e-05, 2.5513e-02, 3.1738e-02, 2.6489e-02, + 3.9673e-03, 2.0386e-02, 3.1250e-02, -4.4005e-08, 1.9531e-02, + 1.3672e-02, 2.8133e-05, 4.5410e-02, 1.1902e-02, 3.6774e-03, + 1.4282e-02, 7.1716e-04, 5.0049e-03, 2.3071e-02, 1.2756e-02, + -1.3447e-04, 1.2695e-02, 3.9795e-02, 2.0264e-02, 4.8096e-02, + 4.0283e-02, 3.1128e-02, 1.8188e-02, 3.4094e-05, 4.0054e-04, + 3.0518e-04, 1.0925e-02, 1.9775e-02, 1.8921e-02, 1.9165e-02, + 3.7689e-03, 1.3672e-02, 3.6621e-02, 4.9414e-01, -8.8289e-07, + 2.5146e-02, 6.9885e-03, 1.5869e-02, 1.6556e-03, 1.1230e-01, + 2.8312e-06, 2.1338e-05, 3.6478e-05, 1.1719e-02, 1.5076e-02, + 1.5747e-02, 2.3804e-02, 3.1738e-02], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.v_proj.weight': tensor([[-0.0183, 0.0031, 0.0049, ..., 0.0004, -0.0172, -0.0260], + [-0.0047, 0.0093, 0.0052, ..., 0.0046, 0.0053, 0.0051], + [-0.0417, 0.0022, -0.0008, ..., -0.0071, -0.0254, -0.0123], + ..., + [-0.0391, 0.0014, -0.0173, ..., 0.0022, 0.0209, 0.0052], + [ 0.0125, 0.0042, -0.0064, ..., -0.0048, 0.0092, 0.0037], + [ 0.0103, 0.0024, -0.0087, ..., -0.0036, -0.0029, 0.0072]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.output_proj.weight': tensor([[-6.5308e-03, -1.1597e-02, 2.0142e-02, ..., 1.0864e-02, + 1.2268e-02, 5.9509e-03], + [-5.8899e-03, -6.9885e-03, -4.7607e-03, ..., -2.9755e-03, + 5.1270e-03, 1.6724e-02], + [-2.5177e-03, 4.0588e-03, -6.8054e-03, ..., -2.5787e-03, + -2.1210e-03, 3.6926e-03], + ..., + [-3.4790e-03, 1.4954e-03, 6.2561e-03, ..., -6.8054e-03, + 6.3477e-03, 3.5553e-03], + [ 4.3640e-03, 4.1504e-03, 3.9368e-03, ..., -3.3203e-02, + -3.5706e-03, 3.3447e-02], + [ 1.1475e-02, 7.6294e-05, -1.5503e-02, ..., -1.7944e-02, + 3.6011e-03, -4.4556e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_img.weight': tensor([[ 0.0082, -0.0017, 0.0430, ..., -0.0327, -0.0052, 0.0155], + [-0.0278, 0.0280, 0.0249, ..., -0.0072, -0.0378, 0.0271], + [ 0.0134, -0.0121, 0.0312, ..., -0.0077, 0.0253, -0.0113], + ..., + [ 0.0245, -0.0150, 0.0041, ..., -0.0452, 0.0216, -0.0209], + [-0.0081, 0.0069, -0.0471, ..., -0.0106, 0.0157, -0.0082], + [ 0.0062, 0.0204, -0.0288, ..., -0.0237, 0.0227, 0.0197]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.v_img.weight': tensor([[ 0.0369, 0.0254, 0.0269, ..., 0.0059, -0.0122, -0.0117], + [ 0.0137, 0.0060, -0.0188, ..., 0.0157, 0.0082, -0.0052], + [-0.0306, 0.0012, -0.0272, ..., 0.0100, 0.0337, -0.0050], + ..., + [-0.0040, -0.0046, 0.0044, ..., 0.0209, -0.0107, 0.0071], + [ 0.0193, 0.0210, 0.0123, ..., -0.0146, 0.0118, 0.0187], + [-0.0251, 0.0366, -0.0305, ..., 0.0109, -0.0131, -0.0291]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img.weight': tensor([[ 0.0045, -0.0057, 0.0173, ..., 0.0081, 0.0356, 0.0217], + [-0.0498, 0.0062, 0.0342, ..., 0.0160, 0.0265, -0.0332], + [ 0.0104, 0.0033, -0.0139, ..., -0.0042, 0.0269, -0.0049], + ..., + [ 0.0010, 0.0126, 0.0079, ..., -0.0266, -0.0508, 0.0270], + [-0.0300, 0.0073, -0.0162, ..., -0.0057, -0.0073, -0.0120], + [ 0.0254, 0.0129, -0.0425, ..., -0.0264, 0.0139, 0.0645]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.1.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.1.mlp.layer1.weight': tensor([[-0.0029, 0.0050, -0.0092, ..., 0.0106, 0.0013, -0.0006], + [-0.0026, 0.0053, 0.0047, ..., 0.0059, 0.0116, -0.0123], + [-0.0070, 0.0026, 0.0059, ..., -0.0065, 0.0070, 0.0033], + ..., + [ 0.0194, -0.0036, 0.0075, ..., 0.0078, -0.0006, 0.0038], + [-0.0242, -0.0052, -0.0021, ..., 0.0217, -0.0055, -0.0046], + [ 0.0086, -0.0250, 0.0101, ..., -0.0140, -0.0094, -0.0160]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.mlp.layer2.weight': tensor([[-0.0098, 0.0081, 0.0093, ..., 0.0049, -0.0040, 0.0064], + [-0.0070, 0.0113, 0.0073, ..., -0.0020, 0.0010, -0.0175], + [ 0.0007, 0.0038, -0.0064, ..., 0.0023, 0.0041, -0.0114], + ..., + [-0.0067, -0.0072, 0.0131, ..., -0.0116, -0.0244, 0.0087], + [ 0.0146, 0.0072, -0.0046, ..., -0.0078, 0.0026, -0.0067], + [-0.0046, 0.0062, 0.0068, ..., 0.0039, -0.0120, -0.0143]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_self_attn.1.weight': tensor([[-0.0405, -0.0396, -0.0391, ..., 0.0050, 0.0053, 0.0052], + [ 0.0327, 0.0315, 0.0306, ..., -0.0037, -0.0036, -0.0037], + [-0.0417, -0.0398, -0.0396, ..., 0.0050, 0.0051, 0.0051], + ..., + [-0.0447, -0.0435, -0.0422, ..., 0.0034, 0.0036, 0.0037], + [-0.0386, -0.0381, -0.0371, ..., 0.0037, 0.0039, 0.0039], + [-0.0228, -0.0220, -0.0210, ..., 0.0012, 0.0012, 0.0013]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_self_attn.2.weight': tensor([[ 0.0043, -0.0006, 0.0045, ..., 0.0017, 0.0025, 0.0002], + [-0.0104, 0.0042, -0.0062, ..., -0.0077, -0.0066, -0.0042], + [-0.0049, 0.0018, -0.0061, ..., -0.0084, -0.0084, -0.0020], + ..., + [-0.0220, 0.0037, -0.0267, ..., -0.0286, -0.0270, -0.0038], + [-0.0124, -0.0055, -0.0131, ..., -0.0153, -0.0137, 0.0064], + [ 0.0064, 0.0004, 0.0054, ..., 0.0051, 0.0044, -0.0004]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9570e-02, 5.7861e-02, 5.7129e-02, ..., -1.2159e-04, + -1.6308e-04, -1.1539e-04], + [-4.0283e-03, -3.9978e-03, -3.9978e-03, ..., -4.1008e-05, + -1.1444e-04, -2.2888e-04], + [-1.8555e-02, -1.7944e-02, -1.7090e-02, ..., -6.9809e-04, + -7.0572e-04, -7.1716e-04], + ..., + [-2.9053e-02, -2.8931e-02, -2.8076e-02, ..., -4.1771e-04, + -3.7384e-04, -3.3569e-04], + [-3.3203e-02, -3.2227e-02, -3.0884e-02, ..., -1.8311e-04, + -1.3638e-04, -1.9073e-04], + [ 5.8838e-02, 5.7129e-02, 5.6396e-02, ..., -2.0504e-04, + -8.8692e-05, -2.3460e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_cross_attn.2.weight': tensor([[-0.0089, 0.0040, -0.0204, ..., 0.0052, 0.0135, -0.0127], + [ 0.0073, 0.0063, -0.0064, ..., 0.0061, -0.0039, 0.0053], + [ 0.0121, 0.0049, -0.0063, ..., 0.0080, -0.0047, 0.0102], + ..., + [ 0.0491, 0.0076, -0.0053, ..., 0.0062, -0.0149, 0.0476], + [ 0.0205, -0.0086, 0.0053, ..., -0.0145, -0.0229, 0.0243], + [-0.0029, 0.0096, -0.0164, ..., 0.0114, 0.0080, -0.0063]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_mlp.1.weight': tensor([[-0.0078, -0.0078, -0.0078, ..., 0.0003, 0.0003, 0.0003], + [ 0.0378, 0.0369, 0.0366, ..., -0.0038, -0.0038, -0.0037], + [ 0.0408, 0.0400, 0.0396, ..., -0.0038, -0.0038, -0.0039], + ..., + [ 0.0262, 0.0253, 0.0254, ..., -0.0021, -0.0021, -0.0020], + [ 0.0057, 0.0063, 0.0064, ..., 0.0014, 0.0014, 0.0013], + [ 0.0166, 0.0170, 0.0167, ..., -0.0025, -0.0025, -0.0026]], + dtype=torch.bfloat16), 'net.control_blocks_3.1.adaln_modulation_mlp.2.weight': tensor([[-3.2654e-03, 6.0120e-03, 4.9133e-03, ..., 4.3640e-03, + 3.2959e-03, 3.2959e-03], + [ 7.0190e-03, 4.7607e-03, 2.6093e-03, ..., -1.2302e-04, + -7.4158e-03, 3.6240e-05], + [ 3.5553e-03, -4.0894e-03, -5.2490e-03, ..., -3.2806e-03, + -3.6621e-03, -3.2501e-03], + ..., + [ 4.0245e-04, 4.3335e-03, -4.7607e-03, ..., -5.8289e-03, + 3.7994e-03, 3.7537e-03], + [ 7.9956e-03, 1.6968e-02, 1.3916e-02, ..., -4.1504e-03, + -7.9956e-03, -3.8147e-03], + [-4.6387e-03, -4.7913e-03, 2.6855e-03, ..., 4.2114e-03, + 4.6387e-03, 4.2419e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.after_proj.weight': tensor([[ 8.1635e-04, 1.4099e-02, 9.2163e-03, ..., -7.5073e-03, + 9.3994e-03, -5.6458e-03], + [-5.2490e-03, -1.0864e-02, -5.1498e-05, ..., 1.3306e-02, + -3.2959e-03, -6.2866e-03], + [ 3.9673e-03, -7.7820e-03, -8.8120e-04, ..., -5.7068e-03, + -1.7944e-02, 6.0120e-03], + ..., + [-5.3101e-03, 6.7749e-03, 2.5757e-02, ..., -2.0020e-02, + -7.4768e-03, -7.4768e-03], + [-1.1902e-02, -5.5237e-03, 1.6632e-03, ..., 1.1230e-02, + 1.0254e-02, -6.3782e-03], + [-4.4556e-03, 7.5378e-03, -4.5776e-03, ..., -6.2561e-03, + -5.9814e-03, 3.9673e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.1.after_proj.bias': tensor([ 0.0052, 0.0009, 0.0016, ..., 0.0064, 0.0066, -0.0004], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_proj.weight': tensor([[ 0.0437, -0.0240, -0.0189, ..., 0.0144, -0.0232, 0.0200], + [-0.0255, -0.0161, -0.0065, ..., -0.0056, 0.0101, 0.0036], + [ 0.0311, 0.0232, 0.0126, ..., -0.0045, -0.0199, -0.0047], + ..., + [-0.0085, -0.0048, 0.0018, ..., 0.0041, 0.0041, -0.0118], + [ 0.0167, 0.0259, 0.0025, ..., 0.0038, -0.0092, 0.0112], + [-0.0107, -0.0045, 0.0186, ..., -0.0060, 0.0145, -0.0156]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_norm.weight': tensor([ 1.2344e+00, 3.1836e-01, 4.5117e-01, 2.8711e-01, 1.1172e+00, + 7.7515e-03, 1.1841e-02, -4.2455e-13, 2.0874e-02, 4.4861e-03, + -5.4250e-08, 1.9096e-13, 1.0000e+00, -7.9632e-05, -1.8860e-22, + -2.9419e-02, -6.1035e-03, 8.8501e-03, 1.6235e-02, 1.3916e-02, + 1.5820e-01, -1.4782e-04, 5.0000e-01, 1.3359e+00, 6.4062e-01, + 7.3828e-01, 3.2812e-01, 1.6602e-01, 3.5156e-02, -1.2146e-02, + -3.3875e-03, 5.7459e-05, 5.9509e-03, 1.6785e-04, -6.7749e-03, + 3.3936e-02, 1.1816e-01, -9.1553e-05, 2.4780e-02, 1.1504e-05, + 1.4210e-04, -1.5974e-05, 4.5076e-07, 1.0469e+00, 1.2793e-01, + 1.1875e+00, 6.4062e-01, 1.4355e-01, -4.2969e-02, 1.1108e-02, + 1.0620e-02, 1.6846e-02, 1.3867e-01, -8.9645e-05, 1.6499e-04, + 1.0048e-14, 2.2602e-04, -9.8944e-06, -5.5176e-02, -3.8025e-15, + 3.0640e-02, 7.2718e-06, 5.7983e-03, -3.7231e-03, 5.8594e-01, + 1.1562e+00, 1.1328e+00, 1.1406e+00, 3.1250e-01, -1.2634e-02, + 8.2397e-03, 5.3957e-14, 2.1118e-02, -1.2085e-02, -7.3924e-09, + -5.5067e-14, 4.1504e-02, 2.7008e-03, -3.1764e-20, 6.7188e-01, + 9.1406e-01, -9.2773e-03, -5.6744e-05, 2.6611e-02, 1.1621e-01, + 7.7515e-03, 1.2812e+00, 5.3125e-01, 1.1875e+00, 1.2500e+00, + 2.4121e-01, 4.1992e-01, 4.4922e-02, -1.5869e-02, 4.3335e-03, + 5.0068e-05, -2.5269e-02, -3.8147e-03, 1.7578e-02, 3.5645e-02, + 1.2598e-01, -9.8705e-05, 9.7266e-01, -6.0536e-09, 1.2970e-04, + -1.3471e-05, -1.3279e-10, 6.9531e-01, 1.3750e+00, 1.1953e+00, + 1.2969e+00, 3.3789e-01, 2.5879e-02, 2.0142e-02, 2.1973e-02, + 2.7344e-02, 1.3574e-01, -1.0347e-04, 1.6594e-04, 1.6542e-14, + 1.5068e-04, -1.1206e-05, 1.0469e+00, -4.8572e-15, 3.5889e-02, + 1.0058e-06, 5.9375e-01, 1.0547e+00], dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.self_attn.k_proj.weight': tensor([[-0.0256, 0.0096, 0.0081, ..., 0.0226, -0.0024, 0.0109], + [ 0.0042, 0.0034, 0.0022, ..., -0.0066, 0.0105, -0.0008], + [-0.0291, 0.0134, -0.0076, ..., -0.0049, 0.0238, 0.0005], + ..., + [-0.0023, -0.0036, 0.0055, ..., -0.0068, -0.0052, 0.0079], + [ 0.0024, 0.0055, 0.0126, ..., 0.0192, -0.0034, -0.0075], + [ 0.0066, 0.0018, 0.0126, ..., -0.0008, 0.0061, 0.0229]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.k_norm.weight': tensor([ 1.2812e+00, 7.1875e-01, 4.3555e-01, 2.9297e-01, 1.1172e+00, + 1.2939e-02, 1.9653e-02, 2.4158e-13, 2.1484e-02, 7.6904e-03, + -6.1002e-08, 1.9895e-13, 1.0000e+00, -7.9632e-05, -7.0310e-23, + -7.6904e-03, 2.6855e-03, 8.0078e-02, 1.8066e-02, 3.1586e-03, + 5.7678e-03, -1.0490e-04, 5.3125e-01, 1.2188e+00, 2.1191e-01, + 5.8594e-01, 4.0430e-01, 1.5625e-01, 1.0193e-02, -1.8692e-03, + 4.6997e-03, 4.8161e-05, 4.1199e-03, 4.0283e-02, 4.8218e-03, + -4.3640e-03, 4.3335e-03, -3.9816e-05, 2.9907e-02, 1.1981e-05, + 1.6880e-04, -1.2636e-05, -4.5449e-07, 1.1406e+00, 7.5781e-01, + 1.1406e+00, 4.4141e-01, 2.2070e-01, 1.6846e-02, 5.6641e-02, + -9.8267e-03, 2.8992e-03, 1.1414e-02, -2.9683e-05, 6.2943e-05, + -1.3600e-14, 2.4605e-04, -3.6210e-06, -2.2217e-02, -3.5388e-15, + -3.8452e-03, -4.6492e-06, 3.6621e-03, 1.3733e-02, 6.2891e-01, + 1.0859e+00, 1.1484e+00, 1.1172e+00, 2.8516e-01, 1.8311e-03, + -2.2125e-03, -2.4691e-13, 1.6357e-02, 1.8539e-03, -7.6252e-09, + -4.5075e-14, -1.3184e-02, 2.7008e-03, 3.1764e-20, 6.7188e-01, + 9.1406e-01, 6.5430e-02, 1.5926e-04, 3.8147e-03, 8.8501e-03, + 7.9956e-03, 1.2109e+00, 7.9688e-01, 1.3047e+00, 1.2969e+00, + 2.5146e-02, 4.2188e-01, 1.7578e-02, 2.1362e-02, -3.4027e-03, + 3.6001e-05, 2.7954e-02, 4.5410e-02, 2.3193e-02, -8.1177e-03, + -1.8311e-02, -6.0320e-05, 9.7266e-01, 2.7707e-08, 1.6117e-04, + -9.1195e-06, 7.1907e-12, 4.7070e-01, 1.2188e+00, 1.1484e+00, + 1.4141e+00, 3.3789e-01, -3.8330e-02, 1.0205e-01, 2.1729e-02, + 1.2024e-02, 1.1475e-02, -5.6267e-05, 1.5354e-04, -2.2871e-14, + 1.7929e-04, -6.0499e-06, 1.0469e+00, -5.3291e-15, -1.5625e-02, + 1.2964e-06, 5.9375e-01, 1.0547e+00], dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.self_attn.v_proj.weight': tensor([[ 0.0022, 0.0095, -0.0096, ..., -0.0177, 0.0064, 0.0099], + [-0.0042, -0.0049, -0.0126, ..., -0.0092, 0.0050, 0.0063], + [-0.0137, -0.0091, 0.0016, ..., 0.0320, -0.0079, -0.0100], + ..., + [ 0.0110, 0.0151, -0.0223, ..., -0.0342, -0.0092, -0.0047], + [ 0.0007, 0.0009, 0.0025, ..., 0.0045, 0.0035, -0.0098], + [ 0.0040, -0.0181, -0.0162, ..., 0.0044, -0.0074, -0.0038]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.self_attn.output_proj.weight': tensor([[ 0.0148, 0.0173, -0.0488, ..., 0.0023, -0.0142, 0.0139], + [-0.0038, -0.0066, -0.0146, ..., 0.0210, 0.0125, -0.0125], + [ 0.0003, 0.0044, 0.0043, ..., 0.0037, 0.0005, 0.0061], + ..., + [ 0.0070, 0.0051, -0.0271, ..., 0.0037, 0.0056, 0.0059], + [ 0.0009, 0.0048, 0.0065, ..., 0.0038, 0.0160, -0.0135], + [ 0.0040, 0.0108, 0.0039, ..., -0.0024, 0.0085, -0.0031]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_proj.weight': tensor([[-0.0053, 0.0062, -0.0029, ..., 0.0040, 0.0022, 0.0024], + [ 0.0075, -0.0033, 0.0025, ..., 0.0078, 0.0012, 0.0060], + [ 0.0046, 0.0026, -0.0014, ..., -0.0047, 0.0080, -0.0036], + ..., + [-0.0021, 0.0029, -0.0014, ..., -0.0053, 0.0069, -0.0017], + [-0.0088, -0.0069, 0.0020, ..., 0.0078, -0.0011, 0.0081], + [-0.0026, -0.0013, -0.0044, ..., -0.0045, 0.0035, -0.0061]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_norm.weight': tensor([ 2.8076e-03, 1.4062e-01, 2.1118e-02, 6.1719e-01, 2.1240e-02, + 1.4526e-02, 1.2695e-02, 5.3406e-03, 2.0386e-02, 1.2305e-01, + 1.7212e-02, 5.6641e-02, 2.6758e-01, 2.2217e-02, 1.7969e-01, + 2.2461e-02, 6.3965e-02, 2.4292e-02, 5.4016e-03, 2.3438e-02, + 1.0681e-02, 1.6895e-01, 2.6562e-01, 2.0410e-01, 2.4902e-02, + 1.5820e-01, 5.5420e-02, 1.1475e-02, 3.8867e-01, 1.2695e-02, + 1.5015e-02, 5.0781e-01, 4.6631e-02, 4.3701e-02, 9.1553e-03, + 2.0020e-02, 8.9355e-02, 5.0391e-01, 5.7422e-01, 1.1841e-02, + 1.1414e-02, 1.0059e-01, 5.2795e-03, 5.5469e-01, 2.9297e-02, + 4.3213e-02, 1.1169e-02, 3.9648e-01, 8.3618e-03, 2.6978e-02, + 6.7871e-02, 1.5747e-02, 2.0264e-02, 5.1953e-01, 2.7313e-03, + 2.8076e-02, 4.7119e-02, 6.7871e-02, 1.1719e-02, 2.0752e-02, + 3.1281e-03, 1.4587e-02, 5.2490e-03, 3.6328e-01, 2.3193e-02, + 1.7700e-03, 2.4902e-02, 6.4453e-02, 5.5664e-02, 9.1553e-03, + 9.4604e-03, 6.2109e-01, -3.6716e-05, 6.0547e-01, 2.0386e-02, + 4.1962e-04, 1.4587e-02, 3.1055e-01, 2.9297e-02, 4.5586e-04, + 8.1055e-02, 8.2397e-03, 5.2246e-02, 9.8877e-03, 1.6235e-02, + 8.0566e-03, 1.1719e-02, 5.1025e-02, 2.0996e-01, 1.1536e-02, + 1.0437e-02, 2.3560e-02, 4.2480e-02, 3.3447e-02, 3.0640e-02, + 1.7456e-02, 1.3550e-02, 2.3560e-02, 5.0537e-02, 2.7466e-02, + 5.0049e-02, 1.9897e-02, 6.2561e-03, 5.2246e-02, 5.9766e-01, + 6.0547e-02, 7.2021e-03, 4.6875e-01, 1.5015e-02, 2.8931e-02, + 1.2085e-02, 1.4587e-02, 7.8201e-05, 7.9102e-02, 4.5117e-01, + 2.0264e-02, 2.4048e-02, 5.6641e-01, 1.0254e-01, 3.1128e-02, + 1.5015e-02, 6.9531e-01, 1.4062e-01, 1.6235e-02, 1.1169e-02, + 7.8125e-03, 2.2461e-02, 1.2634e-02], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_proj.weight': tensor([[ 0.0013, -0.0034, 0.0126, ..., -0.0002, -0.0052, 0.0074], + [-0.0046, 0.0003, -0.0056, ..., 0.0067, -0.0096, -0.0072], + [-0.0070, -0.0016, -0.0056, ..., 0.0168, -0.0151, -0.0118], + ..., + [-0.0018, 0.0046, -0.0001, ..., -0.0026, -0.0044, -0.0068], + [ 0.0043, 0.0029, -0.0074, ..., -0.0056, 0.0054, 0.0016], + [-0.0072, 0.0007, 0.0049, ..., 0.0015, 0.0039, 0.0058]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_norm.weight': tensor([ 2.8381e-03, 1.4062e-01, 2.0996e-02, 6.1719e-01, 2.0996e-02, + 1.4526e-02, 1.2817e-02, 5.2185e-03, 2.0630e-02, 1.2305e-01, + 1.7212e-02, 5.5176e-02, 2.6758e-01, 2.2095e-02, 1.7969e-01, + 2.3193e-02, 6.3477e-02, 2.4414e-02, 5.3406e-03, 2.3315e-02, + 1.0986e-02, 1.6992e-01, 2.6758e-01, 2.0410e-01, 2.4780e-02, + 1.5820e-01, 5.4443e-02, 9.3994e-03, 3.9062e-01, 1.2817e-02, + 1.5198e-02, 5.0781e-01, 4.6631e-02, 4.3701e-02, 9.5215e-03, + 1.9653e-02, 8.8867e-02, 5.0391e-01, 5.7422e-01, 1.1780e-02, + 1.1414e-02, 1.0010e-01, 5.2490e-03, 5.5469e-01, 2.9541e-02, + 4.3213e-02, 1.1475e-02, 3.9648e-01, 8.4839e-03, 2.6855e-02, + 6.7871e-02, 1.6357e-02, 2.0386e-02, 5.1953e-01, 3.1128e-03, + 2.7954e-02, 4.6387e-02, 6.7871e-02, 1.1780e-02, 1.9287e-02, + 3.0060e-03, 1.4526e-02, 5.6152e-03, 3.6328e-01, 2.2949e-02, + 1.9302e-03, 2.4414e-02, 6.4453e-02, 5.5664e-02, 9.0942e-03, + 9.8877e-03, 6.2109e-01, -5.1498e-05, 6.0547e-01, 2.0508e-02, + -1.7548e-04, 1.4648e-02, 3.0859e-01, 3.0029e-02, 6.9046e-04, + 8.1543e-02, 8.1787e-03, 5.1514e-02, 9.7656e-03, 1.5747e-02, + 8.1787e-03, 1.2329e-02, 5.1758e-02, 2.0898e-01, 1.1414e-02, + 1.0498e-02, 2.2583e-02, 4.2725e-02, 3.3203e-02, 3.1494e-02, + 1.7456e-02, 1.3611e-02, 2.3560e-02, 5.0537e-02, 2.7344e-02, + 4.9561e-02, 2.0020e-02, 6.8054e-03, 5.2734e-02, 5.9766e-01, + 6.0547e-02, 7.1411e-03, 4.6875e-01, 1.4587e-02, 2.8809e-02, + 1.2756e-02, 1.4587e-02, 8.0109e-05, 7.9102e-02, 4.5117e-01, + 1.9897e-02, 2.4048e-02, 5.6641e-01, 1.0352e-01, 3.1250e-02, + 1.4771e-02, 6.9922e-01, 1.3965e-01, 1.5259e-02, 1.0803e-02, + 7.6599e-03, 2.2339e-02, 1.2390e-02], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.v_proj.weight': tensor([[ 0.0130, -0.0059, -0.0171, ..., 0.0042, -0.0138, -0.0076], + [-0.0179, -0.0020, -0.0033, ..., -0.0018, -0.0227, 0.0190], + [-0.0127, -0.0049, 0.0033, ..., -0.0045, 0.0310, 0.0281], + ..., + [-0.0030, -0.0030, -0.0121, ..., 0.0048, 0.0171, 0.0043], + [ 0.0278, 0.0032, 0.0175, ..., 0.0035, 0.0116, 0.0164], + [ 0.0128, -0.0001, 0.0049, ..., 0.0053, -0.0192, -0.0161]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.output_proj.weight': tensor([[ 0.0125, -0.0001, -0.0154, ..., -0.0190, 0.0078, -0.0012], + [ 0.0089, 0.0038, 0.0125, ..., 0.0066, -0.0090, 0.0024], + [ 0.0059, -0.0033, -0.0007, ..., 0.0007, -0.0078, -0.0057], + ..., + [-0.0114, 0.0038, 0.0079, ..., 0.0052, -0.0232, 0.0073], + [ 0.0097, -0.0165, 0.0049, ..., -0.0265, -0.0133, -0.0056], + [ 0.0042, -0.0167, 0.0177, ..., 0.0317, -0.0025, 0.0086]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_img.weight': tensor([[ 0.0120, 0.0065, 0.0352, ..., 0.0106, -0.0200, 0.0315], + [-0.0182, -0.0466, 0.0298, ..., 0.0280, -0.0188, 0.0113], + [-0.0032, 0.0150, -0.0159, ..., -0.0039, -0.0054, 0.0679], + ..., + [-0.0208, 0.0129, -0.0273, ..., 0.0308, -0.0071, -0.0067], + [-0.0337, -0.0040, 0.0054, ..., -0.0041, 0.0172, 0.0231], + [-0.0255, -0.0040, -0.0039, ..., 0.0408, 0.0430, -0.0126]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.v_img.weight': tensor([[-0.0276, -0.0237, 0.0332, ..., -0.0378, -0.0280, 0.0017], + [ 0.0198, -0.0208, 0.0128, ..., 0.0090, -0.0337, -0.0050], + [ 0.0302, 0.0101, -0.0029, ..., -0.0304, 0.0199, 0.0045], + ..., + [ 0.0171, -0.0276, -0.0073, ..., 0.0231, 0.0182, -0.0040], + [ 0.0221, 0.0388, 0.0056, ..., 0.0023, -0.0488, -0.0361], + [ 0.0058, 0.0118, -0.0055, ..., 0.0042, 0.0304, -0.0459]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img.weight': tensor([[ 2.7344e-02, -1.9908e-05, 1.1841e-02, ..., 1.4725e-03, + 3.4668e-02, 6.1523e-02], + [-2.5757e-02, 2.6703e-03, -1.9653e-02, ..., 1.3916e-02, + -4.6082e-03, 1.0986e-02], + [-8.2397e-03, -1.6846e-02, -4.2725e-02, ..., -5.8594e-02, + 1.5320e-02, 8.5449e-03], + ..., + [ 9.5825e-03, 1.3184e-02, 1.8433e-02, ..., 1.1047e-02, + 8.9111e-03, 5.1514e-02], + [-3.2806e-03, -2.4658e-02, -2.7832e-02, ..., -8.9722e-03, + -3.8818e-02, 8.3618e-03], + [ 1.9043e-02, -2.6733e-02, -2.1973e-02, ..., -4.5654e-02, + -1.5991e-02, 1.6113e-02]], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.2.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.2.mlp.layer1.weight': tensor([[-0.0068, 0.0051, 0.0058, ..., 0.0002, -0.0018, -0.0094], + [-0.0115, 0.0060, 0.0082, ..., 0.0094, -0.0030, 0.0022], + [ 0.0054, -0.0074, 0.0063, ..., 0.0050, -0.0054, -0.0349], + ..., + [-0.0066, -0.0027, 0.0054, ..., 0.0137, 0.0046, -0.0133], + [ 0.0049, 0.0027, -0.0034, ..., -0.0194, -0.0060, -0.0376], + [ 0.0045, 0.0277, 0.0002, ..., -0.0041, 0.0078, -0.0072]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.mlp.layer2.weight': tensor([[ 0.0056, 0.0057, -0.0063, ..., -0.0060, 0.0172, -0.0059], + [-0.0040, -0.0027, -0.0060, ..., 0.0073, -0.0140, -0.0192], + [ 0.0028, 0.0055, 0.0098, ..., 0.0089, 0.0079, -0.0041], + ..., + [-0.0043, 0.0032, -0.0209, ..., -0.0107, 0.0076, -0.0089], + [-0.0117, 0.0106, 0.0123, ..., -0.0101, 0.0023, -0.0024], + [-0.0011, 0.0063, -0.0120, ..., 0.0049, -0.0094, -0.0374]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_self_attn.1.weight': tensor([[ 0.0347, 0.0344, 0.0330, ..., -0.0041, -0.0042, -0.0043], + [-0.0334, -0.0322, -0.0317, ..., 0.0043, 0.0047, 0.0046], + [-0.0349, -0.0344, -0.0332, ..., 0.0045, 0.0048, 0.0048], + ..., + [-0.0393, -0.0381, -0.0374, ..., 0.0022, 0.0023, 0.0024], + [-0.0400, -0.0393, -0.0383, ..., 0.0021, 0.0022, 0.0022], + [ 0.0078, 0.0078, 0.0073, ..., 0.0002, 0.0002, 0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_self_attn.2.weight': tensor([[ 0.0079, -0.0075, -0.0080, ..., -0.0095, -0.0101, 0.0018], + [ 0.0047, -0.0041, -0.0047, ..., -0.0023, -0.0013, 0.0017], + [ 0.0089, -0.0084, -0.0087, ..., -0.0045, -0.0043, -0.0007], + ..., + [ 0.0383, -0.0422, -0.0413, ..., -0.0222, -0.0182, 0.0082], + [ 0.0093, -0.0092, -0.0087, ..., -0.0067, -0.0067, -0.0113], + [-0.0181, 0.0188, 0.0179, ..., 0.0146, 0.0138, 0.0118]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_cross_attn.1.weight': tensor([[ 5.9326e-02, 5.8350e-02, 5.7373e-02, ..., -1.2589e-04, + -5.5313e-05, -2.0790e-04], + [ 6.0547e-02, 5.8838e-02, 5.7129e-02, ..., 3.7575e-04, + 4.8828e-04, 5.1880e-04], + [-4.9561e-02, -4.7852e-02, -4.6631e-02, ..., -1.2934e-05, + 4.2439e-05, -6.6280e-05], + ..., + [ 3.5156e-02, 3.4180e-02, 3.3447e-02, ..., 1.0538e-04, + 3.7432e-05, 6.6757e-05], + [ 6.2988e-02, 6.2012e-02, 6.0791e-02, ..., 5.6076e-04, + 5.9891e-04, 5.9128e-04], + [-5.1758e-02, -5.1270e-02, -5.0049e-02, ..., -3.5286e-04, + -3.9864e-04, -3.5858e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_cross_attn.2.weight': tensor([[ 0.0064, 0.0097, -0.0022, ..., -0.0041, -0.0103, 0.0048], + [-0.0140, 0.0099, 0.0109, ..., -0.0084, -0.0104, 0.0062], + [ 0.0036, 0.0033, -0.0029, ..., 0.0036, -0.0033, 0.0033], + ..., + [ 0.0182, 0.0115, -0.0118, ..., 0.0065, 0.0120, 0.0056], + [ 0.0114, 0.0291, -0.0128, ..., 0.0134, 0.0383, -0.0232], + [-0.0200, 0.0195, 0.0177, ..., -0.0133, -0.0099, 0.0231]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_mlp.1.weight': tensor([[-0.0757, -0.0742, -0.0718, ..., 0.0029, 0.0032, 0.0031], + [-0.0258, -0.0243, -0.0233, ..., 0.0006, 0.0007, 0.0006], + [ 0.0559, 0.0542, 0.0530, ..., -0.0017, -0.0017, -0.0017], + ..., + [ 0.0430, 0.0420, 0.0415, ..., -0.0019, -0.0018, -0.0016], + [ 0.0347, 0.0330, 0.0317, ..., -0.0009, -0.0008, -0.0009], + [-0.0520, -0.0505, -0.0496, ..., 0.0022, 0.0022, 0.0021]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.adaln_modulation_mlp.2.weight': tensor([[-0.0266, -0.0076, 0.0162, ..., 0.0093, 0.0091, -0.0166], + [ 0.0042, -0.0098, -0.0094, ..., -0.0092, 0.0098, 0.0092], + [-0.0038, -0.0087, -0.0049, ..., -0.0069, 0.0090, 0.0049], + ..., + [ 0.0608, -0.0090, -0.0262, ..., -0.0195, 0.0083, 0.0270], + [ 0.0187, -0.0172, -0.0186, ..., -0.0171, 0.0171, 0.0179], + [ 0.0054, -0.0179, -0.0103, ..., -0.0129, 0.0184, 0.0098]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.after_proj.weight': tensor([[-0.0054, 0.0061, -0.0078, ..., -0.0073, -0.0055, -0.0062], + [-0.0144, -0.0081, 0.0093, ..., 0.0078, -0.0068, 0.0075], + [ 0.0061, 0.0078, 0.0070, ..., -0.0076, 0.0082, -0.0051], + ..., + [ 0.0074, -0.0090, 0.0143, ..., 0.0154, 0.0192, 0.0157], + [ 0.0014, 0.0089, 0.0025, ..., -0.0200, -0.0118, 0.0017], + [-0.0080, -0.0099, -0.0138, ..., -0.0076, -0.0044, -0.0057]], + dtype=torch.bfloat16), 'net.control_blocks_3.2.after_proj.bias': tensor([ 0.0055, -0.0029, 0.0007, ..., 0.0034, 0.0079, 0.0055], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_proj.weight': tensor([[-0.0192, 0.0198, 0.0232, ..., -0.0031, -0.0097, -0.0026], + [-0.0106, -0.0056, -0.0008, ..., 0.0075, -0.0066, -0.0070], + [-0.0182, -0.0057, 0.0047, ..., -0.0069, -0.0079, 0.0070], + ..., + [-0.0043, 0.0033, -0.0081, ..., -0.0080, 0.0104, 0.0029], + [ 0.0066, -0.0008, -0.0037, ..., -0.0289, -0.0199, -0.0289], + [ 0.0036, -0.0161, 0.0070, ..., 0.0107, -0.0027, -0.0008]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_norm.weight': tensor([ 9.3750e-01, 1.4609e+00, 1.3750e+00, 1.3594e+00, 1.3574e-01, + 5.8350e-02, 1.6479e-02, -2.1770e-08, -2.4435e-33, 3.0823e-03, + 5.4016e-03, 1.0840e-01, -6.6223e-03, -1.0315e-02, 1.9989e-03, + 1.4687e-04, 5.9326e-02, 9.1406e-01, 1.0156e+00, 3.9291e-04, + 1.4400e-04, 1.1406e+00, 1.6562e+00, 1.6406e+00, 1.0000e+00, + 8.8672e-01, 1.2969e+00, -1.0376e-02, 4.2439e-05, 6.1989e-05, + -9.3994e-03, 2.6131e-04, 2.0874e-02, 1.0078e+00, 9.3384e-03, + 1.0312e+00, 2.8076e-03, 8.0109e-05, 2.3438e-02, -8.1177e-03, + 5.2527e-07, 1.2207e-04, 1.9908e-05, 1.6797e+00, 1.2266e+00, + 1.2734e+00, 1.3438e+00, 1.4531e+00, 6.7871e-02, 1.7452e-04, + 2.9053e-02, 2.6172e-01, 1.0469e+00, 9.0820e-02, 5.1737e-05, + 1.0596e-01, 2.3804e-03, 1.1621e-01, 2.5630e-05, 6.9580e-03, + 1.9372e-06, -1.8845e-03, 3.7231e-03, 1.0938e+00, 1.5234e+00, + 6.7969e-01, 8.3594e-01, 1.1406e+00, 6.0303e-02, -3.1738e-02, + 3.4637e-03, -4.9127e-08, -3.9482e-33, 2.0752e-02, 1.4709e-02, + 2.5482e-03, 3.1281e-04, -2.6093e-03, -1.6724e-02, 1.3062e-02, + 5.8594e-02, -2.8076e-03, 1.1353e-02, 2.4986e-04, 1.5640e-04, + -6.5002e-03, 6.8359e-01, 6.7578e-01, 1.3281e+00, 1.3438e+00, + 7.0312e-01, 1.0681e-03, -3.3140e-05, 3.1471e-05, 4.6387e-03, + -4.5967e-04, 2.9297e-02, 1.6113e-01, 9.0332e-03, 6.4453e-02, + 1.9932e-04, 3.4809e-05, 1.2207e-01, -7.6771e-05, 5.1498e-05, + -2.6226e-05, 5.1260e-05, 4.9805e-01, 1.4453e+00, 1.3047e+00, + 8.4375e-01, 8.4375e-01, 7.3730e-02, -1.7166e-03, 2.7832e-02, + 1.0547e+00, 5.9570e-02, 9.8047e-01, 4.8161e-05, 2.8839e-03, + 2.6245e-02, 5.2490e-02, 1.2815e-05, 5.5695e-04, 9.3937e-05, + 2.3804e-03, 1.0156e+00, -1.0071e-02], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.self_attn.k_proj.weight': tensor([[-0.0176, -0.0067, -0.0075, ..., -0.0074, 0.0067, -0.0098], + [-0.0050, -0.0165, 0.0178, ..., -0.0085, -0.0135, 0.0146], + [-0.0064, 0.0046, 0.0004, ..., -0.0054, -0.0054, 0.0057], + ..., + [-0.0082, 0.0190, -0.0006, ..., 0.0033, 0.0077, 0.0007], + [ 0.0146, -0.0021, 0.0188, ..., -0.0072, 0.0193, -0.0062], + [-0.0028, -0.0042, 0.0037, ..., -0.0037, 0.0043, 0.0022]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.k_norm.weight': tensor([ 7.0312e-01, 1.4141e+00, 1.3672e+00, 1.3047e+00, 1.3672e-01, + 6.5918e-02, 1.1108e-02, -1.8976e-08, -1.8296e-33, -6.6376e-04, + 5.2185e-03, 1.0938e-01, -1.0681e-02, 3.4668e-02, -1.7929e-03, + 5.9843e-05, 3.9673e-03, 9.1406e-01, 1.0156e+00, 4.6921e-04, + 8.9645e-05, 1.1406e+00, 1.4844e+00, 1.4844e+00, 1.6235e-02, + 2.8516e-01, 1.3281e+00, 1.9302e-03, 4.5538e-05, -5.1498e-05, + -8.7280e-03, 5.8746e-04, -1.9287e-02, 1.0000e+00, 6.2561e-03, + 1.0234e+00, 2.6855e-03, 8.8692e-05, -2.5146e-02, -8.1787e-03, + -1.0505e-06, -1.2589e-04, 1.0550e-05, 1.3203e+00, 5.5859e-01, + 1.2031e+00, 1.3984e+00, 1.4297e+00, -4.9438e-03, -7.7820e-04, + 2.4719e-03, -2.2095e-02, 1.0469e+00, 6.2256e-02, -3.4809e-05, + 1.0596e-01, -1.2970e-03, 1.1621e-01, -7.5400e-06, 7.0801e-02, + -1.7583e-06, 2.1606e-02, -2.1484e-02, 1.0938e+00, 1.5625e+00, + 1.0312e+00, 7.8125e-01, 1.1484e+00, 4.9072e-02, -1.9775e-02, + -1.8799e-02, 5.2620e-08, 4.7426e-33, 2.1240e-02, 2.5391e-02, + 3.3936e-02, 7.2098e-04, 2.7344e-02, -3.0151e-02, -1.3123e-02, + -2.6855e-02, 2.8381e-03, 2.7618e-03, 3.1853e-04, 9.2506e-05, + 5.7678e-03, 1.1875e+00, 1.1484e+00, 1.4297e+00, 1.4141e+00, + 3.0859e-01, 1.1536e-02, -6.0320e-05, -4.3631e-05, 6.0425e-03, + -2.2507e-04, 1.2695e-02, 3.2806e-03, -8.5449e-03, -2.2095e-02, + 9.9182e-05, 5.6267e-05, 1.2158e-01, 3.9339e-05, 6.2943e-05, + 2.7657e-05, 4.7445e-05, 1.2344e+00, 1.6484e+00, 1.2734e+00, + 2.3730e-01, 7.7734e-01, -6.0120e-03, -1.5182e-03, -1.5320e-02, + 1.0547e+00, -2.9053e-02, 9.8438e-01, 4.4823e-05, 7.5989e-03, + 2.6489e-02, 5.2002e-02, 8.0466e-06, 4.1748e-02, 9.4414e-05, + 1.5015e-02, 1.0156e+00, -1.9302e-03], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.self_attn.v_proj.weight': tensor([[ 3.8147e-03, -6.7444e-03, 1.0315e-02, ..., 4.6631e-02, + -9.6893e-04, 2.5330e-03], + [ 1.1963e-02, 2.1973e-02, -8.9722e-03, ..., -2.1240e-02, + -8.8501e-03, -5.7678e-03], + [-2.0294e-03, 1.8677e-02, -4.8218e-03, ..., 1.5625e-02, + -4.4556e-03, 9.4604e-03], + ..., + [ 4.2200e-05, -4.1199e-03, -1.9531e-02, ..., 9.4604e-03, + 1.5381e-02, 3.2349e-03], + [ 2.8229e-03, 1.8433e-02, -5.8899e-03, ..., 3.9978e-03, + 4.2114e-03, -3.4943e-03], + [-9.0332e-03, -8.6060e-03, 5.1880e-03, ..., -6.7139e-03, + -8.2397e-03, -7.4463e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.3.self_attn.output_proj.weight': tensor([[ 0.0072, 0.0190, 0.0056, ..., 0.0101, 0.0192, -0.0095], + [ 0.0275, 0.0043, -0.0049, ..., -0.0041, -0.0270, -0.0094], + [ 0.0039, 0.0334, -0.0115, ..., -0.0043, -0.0165, 0.0060], + ..., + [-0.0381, -0.0177, 0.0038, ..., 0.0108, -0.0082, -0.0183], + [ 0.0012, 0.0125, -0.0266, ..., -0.0176, 0.0238, -0.0033], + [-0.0432, 0.0046, 0.0233, ..., 0.0222, -0.0132, 0.0045]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_proj.weight': tensor([[-0.0003, -0.0081, -0.0045, ..., -0.0159, -0.0014, 0.0152], + [ 0.0070, -0.0017, -0.0037, ..., 0.0017, 0.0044, 0.0084], + [ 0.0084, 0.0048, -0.0092, ..., 0.0050, 0.0126, -0.0039], + ..., + [ 0.0060, 0.0171, 0.0045, ..., 0.0056, -0.0106, 0.0049], + [-0.0036, 0.0073, -0.0050, ..., -0.0056, -0.0114, 0.0037], + [ 0.0004, -0.0054, 0.0040, ..., -0.0049, 0.0120, -0.0042]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_norm.weight': tensor([0.1113, 0.0237, 0.4238, 0.0752, 0.1670, 0.0310, 0.1553, 0.1206, 0.5195, + 0.0447, 0.0654, 0.0161, 0.0518, 0.0674, 0.0640, 0.2021, 0.4570, 0.0344, + 0.7539, 0.0364, 0.0327, 0.1689, 0.0193, 0.0894, 0.0623, 0.4355, 0.5234, + 0.2754, 0.0767, 0.0349, 0.0139, 0.2305, 0.6719, 0.2539, 0.0466, 0.0271, + 0.0216, 0.0967, 0.0184, 0.0383, 0.0527, 0.1201, 0.0229, 0.0439, 0.2080, + 0.0498, 0.3555, 0.0260, 0.0554, 0.0211, 0.0258, 0.4023, 0.5312, 0.3789, + 0.1025, 0.5898, 0.5938, 0.0366, 0.1367, 0.0144, 0.0913, 0.4805, 0.0153, + 0.5273, 0.1699, 0.4160, 0.1016, 0.0498, 0.0732, 0.4277, 0.0281, 0.5508, + 0.0251, 0.0442, 0.1533, 0.0908, 0.0243, 0.1113, 0.6406, 0.0398, 0.0952, + 0.0879, 0.0292, 0.0461, 0.4219, 0.0674, 0.0708, 0.0854, 0.0172, 0.1533, + 0.0432, 0.1787, 0.0471, 0.0703, 0.0391, 0.1201, 0.0114, 0.3555, 0.5391, + 0.0503, 0.0549, 0.3242, 0.0339, 0.1309, 0.0398, 0.2285, 0.0156, 0.5742, + 0.5273, 0.0359, 0.0171, 0.7344, 0.0273, 0.0225, 0.4746, 0.2178, 0.3008, + 0.7227, 0.0369, 0.1001, 0.0801, 0.5625, 0.6055, 0.0199, 0.0188, 0.1914, + 0.0374, 0.0408], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_proj.weight': tensor([[-1.9531e-03, 1.8158e-03, 6.0730e-03, ..., -1.7090e-03, + 1.0925e-02, -1.0864e-02], + [ 5.8289e-03, -6.7711e-05, -1.7776e-03, ..., 1.7548e-03, + -3.8452e-03, 1.2390e-02], + [-1.0437e-02, 7.6904e-03, -1.3428e-02, ..., 1.0803e-02, + 5.3406e-03, 4.2114e-03], + ..., + [ 7.5378e-03, 3.1128e-03, -7.1411e-03, ..., -4.4250e-04, + -1.6724e-02, 5.8594e-03], + [-8.6060e-03, -1.8539e-03, -6.6833e-03, ..., 2.8381e-03, + -4.6692e-03, 1.7853e-03], + [ 9.9487e-03, -1.4038e-03, 5.2795e-03, ..., -7.4768e-03, + 7.4463e-03, -5.7983e-03]], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_norm.weight': tensor([0.1118, 0.0237, 0.4238, 0.0752, 0.1670, 0.0317, 0.1553, 0.1206, 0.5195, + 0.0452, 0.0664, 0.0161, 0.0518, 0.0674, 0.0640, 0.2041, 0.4590, 0.0344, + 0.7500, 0.0361, 0.0325, 0.1699, 0.0194, 0.0889, 0.0618, 0.4316, 0.5234, + 0.2734, 0.0767, 0.0349, 0.0150, 0.2324, 0.6719, 0.2539, 0.0474, 0.0271, + 0.0214, 0.0967, 0.0186, 0.0378, 0.0525, 0.1211, 0.0233, 0.0444, 0.2080, + 0.0498, 0.3555, 0.0262, 0.0557, 0.0208, 0.0262, 0.4023, 0.5312, 0.3789, + 0.1025, 0.5898, 0.5977, 0.0366, 0.1367, 0.0147, 0.0913, 0.4805, 0.0156, + 0.5273, 0.1689, 0.4160, 0.1011, 0.0498, 0.0732, 0.4297, 0.0282, 0.5508, + 0.0249, 0.0439, 0.1533, 0.0903, 0.0243, 0.1118, 0.6406, 0.0398, 0.0942, + 0.0879, 0.0292, 0.0461, 0.4219, 0.0679, 0.0698, 0.0850, 0.0175, 0.1553, + 0.0442, 0.1777, 0.0471, 0.0708, 0.0393, 0.1206, 0.0114, 0.3555, 0.5391, + 0.0498, 0.0552, 0.3242, 0.0339, 0.1309, 0.0398, 0.2285, 0.0156, 0.5742, + 0.5273, 0.0356, 0.0171, 0.7344, 0.0266, 0.0227, 0.4766, 0.2178, 0.3008, + 0.7188, 0.0364, 0.0981, 0.0801, 0.5625, 0.6055, 0.0198, 0.0188, 0.1914, + 0.0386, 0.0410], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.v_proj.weight': tensor([[ 0.0084, -0.0016, -0.0292, ..., 0.0041, 0.0038, -0.0220], + [ 0.0032, 0.0095, -0.0042, ..., 0.0006, -0.0060, -0.0199], + [ 0.0332, -0.0038, -0.0092, ..., -0.0049, -0.0175, -0.0386], + ..., + [-0.0118, 0.0019, 0.0160, ..., -0.0067, -0.0212, 0.0049], + [-0.0311, -0.0029, -0.0271, ..., -0.0032, 0.0128, -0.0242], + [-0.0053, 0.0071, 0.0076, ..., 0.0010, 0.0092, 0.0043]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.output_proj.weight': tensor([[ 0.0048, 0.0059, -0.0087, ..., -0.0060, 0.0027, 0.0042], + [-0.0173, -0.0145, -0.0193, ..., 0.0061, -0.0162, -0.0192], + [ 0.0070, -0.0029, -0.0039, ..., -0.0020, -0.0012, -0.0052], + ..., + [ 0.0082, -0.0060, 0.0193, ..., 0.0190, -0.0111, -0.0039], + [-0.0043, -0.0080, -0.0049, ..., 0.0244, -0.0148, 0.0232], + [-0.0089, -0.0129, 0.0269, ..., -0.0228, 0.0344, 0.0085]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.attn_op._extra_state': tensor([128, 4, 78, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_img.weight': tensor([[ 0.0162, -0.0425, -0.0280, ..., 0.0237, -0.0417, -0.0466], + [-0.0109, 0.0293, 0.0256, ..., -0.0233, -0.0198, -0.0232], + [-0.0121, 0.0049, -0.0066, ..., 0.0266, -0.0278, 0.0076], + ..., + [-0.0217, -0.0410, 0.0488, ..., 0.0013, -0.0027, -0.0255], + [ 0.0015, 0.0066, 0.0024, ..., 0.0300, 0.0084, 0.0205], + [-0.0214, 0.0217, 0.0057, ..., -0.0139, -0.0143, -0.0420]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.v_img.weight': tensor([[-0.0630, 0.0099, 0.0027, ..., -0.0236, -0.0109, -0.0248], + [-0.0081, 0.0033, 0.0293, ..., 0.0066, 0.0087, -0.0383], + [-0.0159, -0.0179, -0.0203, ..., -0.0139, 0.0126, -0.0092], + ..., + [-0.0043, 0.0320, -0.0233, ..., -0.0067, 0.0016, 0.0188], + [ 0.0134, -0.0352, 0.0267, ..., -0.0305, -0.0021, 0.0037], + [-0.0205, 0.0140, -0.0019, ..., -0.0464, -0.0427, 0.0378]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img.weight': tensor([[-0.0080, 0.0184, 0.0199, ..., -0.0082, -0.0210, 0.0540], + [ 0.0088, -0.0019, 0.0165, ..., 0.0066, -0.0214, 0.0183], + [ 0.0057, -0.0537, -0.0282, ..., -0.0109, -0.0145, -0.0016], + ..., + [ 0.0094, -0.0181, -0.0500, ..., -0.0291, -0.0013, 0.0015], + [-0.0140, -0.0422, -0.0151, ..., 0.0186, 0.0221, 0.0510], + [-0.0135, 0.0437, -0.0022, ..., -0.0103, 0.0073, -0.0215]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.q_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.cross_attn.k_img_norm.weight': tensor([1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., 1., + 1., 1.], dtype=torch.bfloat16), 'net.control_blocks_3.3.cross_attn.k_img_norm._extra_state': tensor([128, 4, 125, 148, 46], dtype=torch.uint8), 'net.control_blocks_3.3.mlp.layer1.weight': tensor([[ 0.0108, 0.0060, -0.0140, ..., 0.0150, 0.0309, -0.0071], + [-0.0115, 0.0194, 0.0139, ..., -0.0099, -0.0183, 0.0038], + [-0.0079, 0.0284, -0.0168, ..., 0.0104, -0.0039, 0.0002], + ..., + [ 0.0049, 0.0046, 0.0630, ..., 0.0175, 0.0330, 0.0019], + [ 0.0021, -0.0042, -0.0120, ..., -0.0060, -0.0045, -0.0141], + [ 0.0013, -0.0130, -0.0064, ..., -0.0143, -0.0232, -0.0150]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.mlp.layer2.weight': tensor([[-0.0043, -0.0125, -0.0120, ..., -0.0203, 0.0067, -0.0003], + [-0.0059, 0.0171, 0.0068, ..., 0.0078, 0.0133, 0.0211], + [-0.0038, 0.0016, 0.0134, ..., -0.0035, 0.0135, 0.0177], + ..., + [-0.0152, 0.0171, 0.0059, ..., -0.0117, -0.0147, 0.0052], + [ 0.0100, -0.0058, 0.0141, ..., 0.0024, 0.0131, 0.0106], + [-0.0130, -0.0079, 0.0022, ..., 0.0167, 0.0105, 0.0075]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_self_attn.1.weight': tensor([[-0.0444, -0.0437, -0.0427, ..., 0.0010, 0.0008, 0.0009], + [-0.0413, -0.0403, -0.0393, ..., 0.0027, 0.0028, 0.0027], + [-0.0396, -0.0386, -0.0374, ..., 0.0022, 0.0023, 0.0023], + ..., + [ 0.0498, 0.0481, 0.0469, ..., -0.0036, -0.0037, -0.0037], + [ 0.0398, 0.0386, 0.0371, ..., -0.0031, -0.0031, -0.0030], + [-0.0396, -0.0386, -0.0378, ..., 0.0023, 0.0024, 0.0025]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_self_attn.2.weight': tensor([[ 0.0062, 0.0070, 0.0070, ..., 0.0150, 0.0175, 0.0078], + [-0.0017, -0.0014, -0.0012, ..., -0.0032, -0.0031, -0.0013], + [ 0.0013, -0.0071, -0.0061, ..., 0.0034, 0.0036, -0.0070], + ..., + [-0.0325, -0.0776, -0.0645, ..., 0.0221, 0.0236, -0.0718], + [-0.0105, -0.0099, -0.0103, ..., 0.0039, 0.0040, -0.0096], + [ 0.0082, 0.0050, 0.0067, ..., 0.0162, 0.0187, 0.0062]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_cross_attn.1.weight': tensor([[-8.2520e-02, -8.0566e-02, -7.7637e-02, ..., -1.0586e-04, + -1.2064e-04, -1.1063e-04], + [ 3.3691e-02, 3.2471e-02, 3.1982e-02, ..., 1.9222e-06, + 2.7180e-05, 1.4877e-04], + [ 3.0884e-02, 2.9785e-02, 2.8809e-02, ..., 2.7657e-04, + 3.0136e-04, 2.0695e-04], + ..., + [ 5.0293e-02, 4.8828e-02, 4.8340e-02, ..., -1.7929e-04, + -1.7262e-04, -1.0538e-04], + [ 5.4932e-02, 5.3467e-02, 5.2002e-02, ..., -1.6785e-04, + -1.0729e-04, -1.5450e-04], + [-2.5635e-02, -2.4902e-02, -2.4292e-02, ..., -6.2943e-05, + 9.0122e-05, 1.1253e-04]], dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_cross_attn.2.weight': tensor([[-0.0039, -0.0059, 0.0014, ..., 0.0275, 0.0325, 0.0023], + [ 0.0033, -0.0044, 0.0059, ..., -0.0010, 0.0016, 0.0042], + [ 0.0018, -0.0019, 0.0026, ..., -0.0084, -0.0086, -0.0021], + ..., + [-0.0198, 0.0078, 0.0280, ..., 0.0178, 0.0287, -0.0089], + [-0.0222, 0.0126, 0.0347, ..., -0.0042, -0.0026, -0.0033], + [ 0.0179, -0.0112, -0.0062, ..., -0.0096, -0.0111, 0.0176]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_mlp.1.weight': tensor([[-0.0771, -0.0752, -0.0737, ..., 0.0002, 0.0003, 0.0003], + [ 0.0282, 0.0273, 0.0267, ..., 0.0003, 0.0004, 0.0004], + [-0.0991, -0.0977, -0.0962, ..., -0.0003, -0.0003, -0.0003], + ..., + [-0.0723, -0.0708, -0.0698, ..., 0.0002, 0.0003, 0.0002], + [-0.0540, -0.0527, -0.0518, ..., 0.0003, 0.0001, 0.0001], + [ 0.0703, 0.0693, 0.0679, ..., -0.0002, -0.0002, -0.0002]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.adaln_modulation_mlp.2.weight': tensor([[-0.0166, -0.0019, -0.0118, ..., -0.0186, -0.0221, 0.0200], + [-0.0018, -0.0050, 0.0067, ..., 0.0004, 0.0036, -0.0035], + [-0.0019, 0.0020, -0.0027, ..., -0.0019, -0.0019, 0.0019], + ..., + [-0.0486, -0.0170, -0.0557, ..., -0.0356, -0.0126, 0.0272], + [ 0.0071, -0.0099, 0.0264, ..., 0.0056, 0.0048, -0.0047], + [ 0.0138, 0.0110, 0.0107, ..., 0.0124, 0.0065, -0.0101]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.after_proj.weight': tensor([[ 0.0066, -0.0078, -0.0059, ..., -0.0055, 0.0026, -0.0152], + [ 0.0037, -0.0080, 0.0184, ..., -0.0098, -0.0061, -0.0095], + [ 0.0123, 0.0111, 0.0305, ..., 0.0052, 0.0048, -0.0076], + ..., + [ 0.0073, 0.0186, -0.0070, ..., 0.0038, -0.0143, -0.0076], + [-0.0108, -0.0059, -0.0055, ..., 0.0070, -0.0120, 0.0055], + [-0.0166, 0.0165, -0.0107, ..., -0.0049, 0.0009, 0.0073]], + dtype=torch.bfloat16), 'net.control_blocks_3.3.after_proj.bias': tensor([ 0.0026, -0.0143, -0.0110, ..., 0.0014, -0.0015, 0.0062], + dtype=torch.bfloat16)} +[06-26 07:20:52|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:919:load_state_dict] load model in non-strict mode +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.4.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.5.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.6.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.7.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.8.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.9.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.10.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.11.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.12.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.13.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.14.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.15.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.16.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.17.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.18.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.19.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.20.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.21.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.22.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.23.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.24.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.25.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.26.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key blocks.27.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key t_embedding_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.0.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.1.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.2.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.self_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.self_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.cross_attn.q_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:52|WARNING|cosmos_transfer2/_src/imaginaire/utils/checkpointer.py:451:non_strict_load_model] Skipping key control_blocks_3.3.cross_attn.k_norm._extra_state introduced by TransformerEngine for FP8 in the checkpoint. +[06-26 07:20:53|CRITICAL|cosmos_transfer2/_src/predict2/models/text2world_model_rectified_flow.py:920:load_state_dict] [RANK 0] _IncompatibleKeys(missing_keys=['control_embedder.0.proj.1.weight', 'control_embedder.1.proj.1.weight', 'control_embedder.2.proj.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_0.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_0.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_1.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_1.3._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.0._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.before_proj.bias', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.0._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.1._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.1._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.2._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.2._checkpoint_wrapped_module.after_proj.bias', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.self_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.q_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.k_norm.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.v_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.cross_attn.output_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.mlp.layer2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_self_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_cross_attn.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.1.weight', 'control_blocks_2.3._checkpoint_wrapped_module.adaln_modulation_mlp.2.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.weight', 'control_blocks_2.3._checkpoint_wrapped_module.after_proj.bias'], unexpected_keys=['img_context_proj.0.weight', 'img_context_proj.0.bias', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.4._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.5._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.6._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.7._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.8._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.9._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.10._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.11._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.12._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.13._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.14._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.15._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.16._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.17._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.18._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.19._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.20._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.21._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.22._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.23._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.24._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.25._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.26._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.v_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'blocks.27._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.0._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.1._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.2._checkpoint_wrapped_module.cross_attn.k_img_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.v_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_img.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.q_img_norm.weight', 'control_blocks_3.3._checkpoint_wrapped_module.cross_attn.k_img_norm.weight'], incorrect_shapes=[]) +[06-26 07:20:53|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:735:load_multi_branch_checkpoints] Done loading the control branch 3 checkpoint. +[06-26 07:20:53|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:1015:compile_tokenizer_if_enabled] Tokenizer compilation disabled +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:146:__init__] Found 2 hint keys across all samples +[06-26 07:20:53|WARNING|cosmos_transfer2/inference.py:148:__init__] Loading the multicontrol model. Multicontrol inference is not strictly equal to single control +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:156:__init__] Saved config to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/config.yaml +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:163:generate] Generating 1 samples: ['sunny_weather_lighting'] +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:167:generate] [1/1] Processing sample sunny_weather_lighting +[06-26 07:20:53|DEBUG|cosmos_transfer2/inference.py:186:_generate_sample] InferenceArguments(name='sunny_weather_lighting' prompt_path=None prompt='The video shows an urban intersection captured from a vehicle-mounted camera on a brilliant clear sunny day. A deep blue cloudless sky sits over the scene, warm directional sunlight casts long crisp shadows across the dry asphalt, and the painted lane markings, stop lines and pedestrian crosswalks are bright and high-contrast. Multiple vehicles move through the junction, low-rise buildings, sidewalks, parked cars and street lamps line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, cinematic daylight, high detail.' negative_prompt='The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Overcast, rain, wet road, puddles, fog, night, dark shadows.' seed=2025 guidance=3 video_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4') max_frames=None context_frame_index=None image_context_path=None num_conditional_frames=1 resolution='720' sigma_max=None num_video_frames_per_chunk=93 num_steps=35 show_control_condition=False show_input=False keep_input_resolution=True edge=EdgeConfig(control_path=PosixPath('/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4'), control_weight=1.0, mask_path=None, mask_prompt=None, preset_edge_threshold='medium') depth=None vis=BlurConfig(control_path=None, control_weight=0.2, mask_path=None, mask_prompt=None, preset_blur_strength='medium') seg=None guided_generation_mask=None guided_generation_step_threshold=25 guided_generation_foreground_labels=None) +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:208:_generate_sample] Saved arguments to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.json +[06-26 07:20:53|WARNING|cosmos_transfer2/inference.py:239:_generate_sample] Guardrail checks on prompt are disabled +[06-26 07:20:53|INFO|cosmos_transfer2/inference.py:242:_generate_sample] Processing the following paths: {'edge': '/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4', 'edge_mask': None, 'edge_mask_prompt': None, 'vis': None, 'vis_mask': None, 'vis_mask_prompt': None} +[06-26 07:20:53|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:472:generate_img2world] Loading input video... +[06-26 07:20:53|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4 +[06-26 07:20:55|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 07:20:55|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:492:generate_img2world] Computing prompt text embeddings... +[06-26 07:21:01|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:501:generate_img2world] Computing negative prompt text embeddings... +[06-26 07:21:01|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:513:generate_img2world] Processing image context if available... +[06-26 07:21:01|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:330:read_and_process_image_context] No image context provided. +[06-26 07:21:01|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:525:generate_img2world] Loading control inputs... +[06-26 07:21:01|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:204:read_video_or_image_into_frames_BCTHW] Reading video from /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4 +[06-26 07:21:02|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:243:read_video_or_image_into_frames_BCTHW] Loaded input tensor with shape (1, 3, 120, 720, 1280) value 0, 255 +[06-26 07:21:02|INFO|cosmos_transfer2/_src/transfer2/inference/utils.py:703:read_and_process_control_input] No vis (blur) control input file found, will compute online.. +[06-26 07:21:02|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 1/2 +[06-26 07:21:10|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 584845 +[06-26 07:21:10|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 0 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 07:27:30|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:554:generate_img2world] Generating chunk 2/2 +[06-26 07:27:37|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:640:generate_img2world] Seed: 643774 +[06-26 07:27:38|INFO|cosmos_transfer2/_src/transfer2/models/vid2vid_model_control_vace_rectified_flow.py:239:get_velocity_fn_from_batch] num_conditional_frames: 1 is set by data_batch[NUM_CONDITIONAL_FRAMES_KEY] +[06-26 07:33:18|INFO|cosmos_transfer2/_src/transfer2/inference/inference_pipeline.py:736:generate_img2world] Average time per chunk: 365.6185655920017 +[06-26 07:33:19|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 07:33:20|INFO|cosmos_transfer2/inference.py:311:_generate_sample] edge control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_edge.mp4 +[06-26 07:33:20|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 07:33:21|INFO|cosmos_transfer2/inference.py:311:_generate_sample] vis control video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_vis.mp4 +[06-26 07:33:21|WARNING|cosmos_transfer2/inference.py:335:_generate_sample] Guardrail checks on video are disabled +[06-26 07:33:21|DEBUG|cosmos_transfer2/_src/imaginaire/utils/easy_io/handlers/imageio_video_handler.py:163:dump_to_fileobj] mimsave_kwargs: {'fps': 10, 'quality': 5, 'macro_block_size': 1, 'ffmpeg_params': ['-s', '1280x720'], 'output_params': ['-f', 'mp4']} +[06-26 07:33:22|SUCCESS|cosmos_transfer2/inference.py:343:_generate_sample] Generated video saved to /scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.mp4 diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/spec.json b/multicontrol_cookbook_grid/sunny/weather/lighting/spec.json new file mode 100644 index 0000000000000000000000000000000000000000..6e2b7e8f1cf0fc0a7ac60ec3ff38e24d8bc63984 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/spec.json @@ -0,0 +1,16 @@ +{ + "name": "sunny_weather_lighting", + "prompt": "The video shows an urban intersection captured from a vehicle-mounted camera on a brilliant clear sunny day. A deep blue cloudless sky sits over the scene, warm directional sunlight casts long crisp shadows across the dry asphalt, and the painted lane markings, stop lines and pedestrian crosswalks are bright and high-contrast. Multiple vehicles move through the junction, low-rise buildings, sidewalks, parked cars and street lamps line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, cinematic daylight, high detail.", + "video_path": "/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4", + "guidance": 3, + "num_steps": 35, + "seed": 2025, + "edge": { + "control_path": "/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4", + "control_weight": 1.0 + }, + "vis": { + "control_weight": 0.2 + }, + "negative_prompt": "The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Overcast, rain, wet road, puddles, fog, night, dark shadows." +} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.json b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.json new file mode 100644 index 0000000000000000000000000000000000000000..96f92593a508a9942acae53e2c6b88ab2fbcd05a --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.json @@ -0,0 +1 @@ +{"name":"sunny_weather_lighting","prompt_path":null,"prompt":"The video shows an urban intersection captured from a vehicle-mounted camera on a brilliant clear sunny day. A deep blue cloudless sky sits over the scene, warm directional sunlight casts long crisp shadows across the dry asphalt, and the painted lane markings, stop lines and pedestrian crosswalks are bright and high-contrast. Multiple vehicles move through the junction, low-rise buildings, sidewalks, parked cars and street lamps line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, cinematic daylight, high detail.","negative_prompt":"The video captures a game playing, with bad crappy graphics and cartoonish frames. It represents a recording of old outdated games. The lighting looks very fake. The textures are very raw and basic. The geometries are very primitive. The images are very pixelated and of poor CG quality. There are many subtitles in the footage. Overall, the video is unrealistic at all. Overcast, rain, wet road, puddles, fog, night, dark shadows.","seed":2025,"guidance":3,"video_path":"/scratch/ssaha/cosmos-transfer2.5/SceneDriveBench/output/priority_from_right/run_000/videos/rgb_front.mp4","max_frames":null,"context_frame_index":null,"image_context_path":null,"num_conditional_frames":1,"resolution":"720","sigma_max":null,"num_video_frames_per_chunk":93,"num_steps":35,"show_control_condition":false,"show_input":false,"keep_input_resolution":true,"edge":{"control_path":"/scratch/ssaha/cosmos-transfer2.5/outputs/multicontrol_cookbook_grid/_edges/sunny_edge.mp4","control_weight":1.0,"mask_path":null,"mask_prompt":null,"preset_edge_threshold":"medium"},"depth":null,"vis":{"control_path":null,"control_weight":0.2,"mask_path":null,"mask_prompt":null,"preset_blur_strength":"medium"},"seg":null,"guided_generation_mask":null,"guided_generation_step_threshold":25,"guided_generation_foreground_labels":null} \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.mp4 b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..c7384f4af87552213472d235acf2a8538bfd3792 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:54180d8c7af17e375fdb368b73fc603cabaaf1a8076e8398c22bb98ce59cffcc +size 1759358 diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.txt b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.txt new file mode 100644 index 0000000000000000000000000000000000000000..464b6055f679cb63d0db8e54bd910217c87b403b --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting.txt @@ -0,0 +1 @@ +The video shows an urban intersection captured from a vehicle-mounted camera on a brilliant clear sunny day. A deep blue cloudless sky sits over the scene, warm directional sunlight casts long crisp shadows across the dry asphalt, and the painted lane markings, stop lines and pedestrian crosswalks are bright and high-contrast. Multiple vehicles move through the junction, low-rise buildings, sidewalks, parked cars and street lamps line the streets, and the camera follows the moving ego vehicle smoothly. Photorealistic, cinematic daylight, high detail. \ No newline at end of file diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_edge.mp4 b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_edge.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..517a976ec0a5d8e4a484eb65a5ac5ad8308f733a --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_edge.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57c1f383a1f61fc52a275e2a3d237bee50a45e7901dc709b401226496891cd0c +size 1406752 diff --git a/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_vis.mp4 b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_vis.mp4 new file mode 100644 index 0000000000000000000000000000000000000000..34f0b6e199ee4029ebdfebc39c17fa0c9132a9f0 --- /dev/null +++ b/multicontrol_cookbook_grid/sunny/weather/lighting/sunny_weather_lighting_control_vis.mp4 @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:282799a9e77adfeeb3c049be5c9a739ce7b15b57f4e9e386fb937053bde1d587 +size 293428